[2026-10-17 14:56:50] Registered linear attn model: config=GigaChat35Config, backend=GDNAttnBackend, archs=['GigaChat35ForCausalLM', 'GigaChat35ForCausalLMNextN']
[2026-10-17 14:56:52] Failed to get GPU memory capacity from nvidia-smi. Falling back to torch.cuda.mem_get_info(). Reported total GPU memory per device (MiB): [124610], using min: 124610 MiB.
[2026-10-17 14:56:52] Attention backend not specified. Use flashinfer backend by default.
[2026-10-17 14:56:54] server_args={'model_path': 'Qwen/Qwen3-30B-A3B-Instruct-2507', 'tokenizer_path': 'Qwen/Qwen3-30B-A3B-Instruct-2507', 'tokenizer_mode': 'auto', 'tokenizer_backend': 'huggingface', 'tokenizer_worker_num': 1, 'detokenizer_worker_num': 1, 'skip_tokenizer_init': False, 'load_format': 'auto', 'model_loader_extra_config': '{}', 'trust_remote_code': False, 'context_length': None, 'is_embedding': False, 'enable_multimodal': None, 'revision': None, 'model_impl': 'auto', 'model_config_parser': 'auto', 'json_model_override_args': '{}', 'dtype': 'auto', 'quantization': None, 'quantization_param_path': None, 'kv_cache_dtype': 'auto', 'enable_fp32_lm_head': False, 'modelopt_quant': None, 'modelopt_checkpoint_restore_path': None, 'modelopt_checkpoint_save_path': None, 'modelopt_export_path': None, 'quantize_and_serve': False, 'rl_quant_profile': None, 'enable_tf32_matmul': False, 'mem_fraction_static': 0.7, 'max_running_requests': None, 'max_queued_requests': None, 'max_total_tokens': 4096, 'chunked_prefill_size': 8192, 'prefill_decode_interval': 0, 'enable_dynamic_chunking': False, 'max_prefill_tokens': 16384, 'prefill_max_requests': None, 'schedule_policy': 'fcfs', 'enable_priority_scheduling': False, 'disable_priority_preemption': False, 'default_priority_value': None, 'abort_on_priority_when_disabled': False, 'schedule_low_priority_values_first': False, 'priority_scheduling_preemption_threshold': 10, 'retraction_policy': 'length', 'schedule_conservativeness': 1.0, 'page_size': 1, 'c128_page_size': 16, 'swa_full_tokens_ratio': 0.8, 'disable_hybrid_swa_memory': False, 'radix_eviction_policy': 'lru', 'radix_eviction_policy_config': None, 'prefill_only_disable_kv_cache': False, 'disable_radix_cache': False, 'enable_page_major_kv_layout': False, 'enable_unified_memory': False, 'disable_chunked_prefix_cache': False, 'disable_overlap_schedule': False, 'num_continuous_decode_steps': 1, 'scheduler_recv_interval': 1, 'enable_mixed_chunk': False, 'nccl_port': None, 'dist_timeout': None, 'dist_init_addr': None, 'gated_launch_port': None, 'nnodes': 1, 'node_rank': 0, 'tp_size': 1, 'dcp_size': 1, 'pp_size': 1, 'pp_max_micro_batch_size': None, 'pp_async_batch_depth': 0, 'dp_size': 1, 'attn_dp_size': 1, 'load_balance_method': 'round_robin', 'attn_cp_size': 1, 'moe_dp_size': 1, 'dwdp_size': 1, 'dcp_comm_backend': 'ag_rs', 'dcp_replicate_q_proj': None, 'enable_prefill_cp': False, 'cp_strategy': None, 'enable_dsa_cache_layer_split': False, 'enable_cp_decode_attn_tp': False, 'enable_kv_cache_sharding': False, 'enable_dp_attention': False, 'enable_dp_attention_local_control_broadcast': False, 'enable_dp_lm_head': False, 'enable_tp_lm_head_all_to_all': False, 'enable_attn_tp_input_scattered': False, 'enable_shared_experts_attn_tp': False, 'shared_experts_tp_size': None, 'enable_dense_mlp_attn_tp': False, 'enable_layernorm_sp': False, 'disable_attn_tp_gather': False, 'enable_p2p_check': False, 'device': 'cuda', 'base_gpu_id': 0, 'gpu_id_step': 1, 'random_seed': 891134967, 'mlx_enable_sampling': False, 'watchdog_timeout': 300, 'soft_watchdog_timeout': None, 'sleep_on_idle': False, 'use_ray': False, 'custom_sigquit_handler': None, 'numa_node': None, 'gc_threshold': None, 'host': '127.0.0.1', 'port': 30000, 'fastapi_root_path': '', 'smg_grpc_mode': False, 'grpc_mode': False, 'grpc_port': None, 'grpc_worker_threads': 4, 'sidecar': None, 'sidecar_args': None, 'skip_server_warmup': False, 'warmups': None, 'enable_http2': False, 'http2_max_concurrent_streams': 200, 'http2_initial_connection_window_size': 1048576, 'ssl_keyfile': None, 'ssl_certfile': None, 'ssl_ca_certs': None, 'ssl_keyfile_password': None, 'enable_ssl_refresh': False, 'api_key': None, 'admin_api_key': None, 'served_model_name': 'Qwen/Qwen3-30B-A3B-Instruct-2507', 'weight_version': 'default', 'chat_template': None, 'trust_request_chat_template': False, 'hf_chat_template_name': None, 'completion_template': None, 'file_storage_path': None, 'enable_cache_report': True, 'return_input_ids': False, 'return_output_ids': False, 'reasoning_parser': None, 'default_chat_template_kwargs': None, 'strip_thinking_cache': False, 'enable_strict_thinking': False, 'tool_call_parser': None, 'tool_server': None, 'sampling_defaults': 'model', 'asr_max_buffer_seconds': 60, 'asr_max_concurrent_sessions': 32, 'preferred_sampling_params': None, 'allow_auto_truncate': False, 'stream_interval': 1, 'batch_notify_size': 16, 'stream_response_default_include_usage': False, 'incremental_streaming_output': False, 'enable_streaming_session': False, 'enable_session_radix_cache': False, 'log_level': 'info', 'log_level_http': None, 'log_requests': False, 'log_requests_level': 2, 'log_requests_format': 'text', 'log_requests_target': None, 'uvicorn_access_log_exclude_prefixes': [], 'crash_dump_folder': None, 'show_time_cost': False, 'enable_metrics': True, 'smg_http_sidecar_port': None, 'enable_mfu_metrics': False, 'enable_metrics_for_all_schedulers': False, 'load_snapshot_publish_interval': 15, 'tokenizer_metrics_custom_labels_header': 'x-custom-labels', 'tokenizer_metrics_allowed_custom_labels': None, 'extra_metric_labels': None, 'bucket_time_to_first_token': None, 'bucket_inter_token_latency': None, 'bucket_e2e_request_latency': None, 'prompt_tokens_buckets': None, 'generation_tokens_buckets': None, 'gc_warning_threshold_secs': 0.0, 'decode_log_interval': 40, 'enable_request_time_stats_logging': False, 'kv_events_config': None, 'load_publish_endpoint': None, 'enable_forward_pass_metrics': False, 'forward_pass_metrics_worker_id': '', 'forward_pass_metrics_ipc_name': None, 'enable_trace': False, 'trace_modules': 'request', 'otlp_traces_endpoint': 'localhost:4317', 'export_metrics_to_file': False, 'export_metrics_to_file_dir': None, 'stat_loggers': None, 'constrained_json_whitespace_pattern': None, 'constrained_json_disable_any_whitespace': False, 'constrained_json_max_whitespace_cnt': None, 'attention_backend': 'flashinfer', 'decode_attention_backend': None, 'enable_lean_attention': None, 'prefill_attention_backend': None, 'sampling_backend': 'flashinfer', 'grammar_backend': 'xgrammar', 'radix_cache_backend': None, 'mm_attention_backend': None, 'fp8_gemm_runner_backend': 'auto', 'fp4_gemm_runner_backend': 'auto', 'bf16_gemm_backend': 'auto', 'dsa_prefill_backend': None, 'dsv4_prefill_backend': 'auto', 'dsa_decode_backend': None, 'dsa_paged_mqa_logits_backend': 'auto', 'dsa_topk_backend': 'sgl-kernel', 'disable_flashinfer_autotune': False, 'flashinfer_autotune_skip_ops': None, 'mamba_backend': 'triton', 'cuda_graph_config': {'decode': {'backend': 'full', 'max_bs': 256, 'bs': [1, 2, 4, 8, 12, 16, 24, 32, 40, 48, 56, 64, 72, 80, 88, 96, 104, 112, 120, 128, 136, 144, 152, 160, 168, 176, 184, 192, 200, 208, 216, 224, 232, 240, 248, 256], 'tc_compiler': 'eager', 'max_context_size': None, 'full_prefill_max_req': None, 'full_prefill_prefix_chunk_tokens': None, 'max_seq_len': None}, 'prefill': {'backend': 'breakable', 'max_bs': 4096, 'bs': [4, 8, 12, 16, 20, 24, 28, 32, 48, 64, 80, 96, 112, 128, 144, 160, 176, 192, 208, 224, 240, 256, 288, 320, 352, 384, 416, 448, 480, 512, 576, 640, 704, 768, 832, 896, 960, 1024, 1280, 1536, 1792, 2048, 2304, 2560, 2816, 3072, 3328, 3584, 3840, 4096], 'tc_compiler': 'eager', 'max_context_size': None, 'full_prefill_max_req': None, 'full_prefill_prefix_chunk_tokens': None, 'max_seq_len': None}}, 'cuda_graph_backend_decode': None, 'cuda_graph_backend_prefill': None, 'cuda_graph_max_bs_decode': None, 'cuda_graph_max_bs_prefill': None, 'cuda_graph_bs_decode': None, 'cuda_graph_bs_prefill': None, 'cuda_graph_prefill_max_context': None, 'disable_prefill_cuda_graph': False, 'disable_decode_cuda_graph': False, 'disable_cuda_graph': False, 'disable_cuda_graph_padding': False, 'enable_profile_cuda_graph': False, 'enable_cudagraph_gc': False, 'debug_cuda_graph': False, 'enable_layerwise_nvtx_marker': False, 'enable_nccl_nvls': False, 'enable_symm_mem': False, 'triton_attention_reduce_in_fp32': False, 'triton_attention_num_kv_splits': 8, 'triton_attention_split_tile_size': None, 'flashinfer_mla_disable_ragged': False, 'enable_fused_qk_norm_rope': False, 'enable_precise_embedding_interpolation': False, 'enable_fused_moe_sum_all_reduce': False, 'enable_deepseek_v4_fp4_indexer': False, 'disable_custom_all_reduce': False, 'enable_mscclpp': False, 'enable_torch_symm_mem': False, 'enable_scattered_sconv': False, 'pre_warm_nccl': False, 'boundary_reduction': 'rs+rsv', 'enable_quant_communications': False, 'enable_flashinfer_allreduce_fusion': False, 'enforce_disable_flashinfer_allreduce_fusion': False, 'flashinfer_allreduce_fusion_backend': None, 'enable_aiter_allreduce_fusion': False, 'enable_torch_compile': False, 'enable_torch_compile_debug_mode': False, 'torch_compile_max_bs': 32, 'speculative_algorithm': None, 'uno_lora_path': None, 'speculative_draft_model_path': None, 'speculative_draft_model_revision': None, 'speculative_draft_load_format': None, 'speculative_num_steps': None, 'speculative_eagle_topk': None, 'speculative_num_draft_tokens': None, 'speculative_dflash_block_size': None, 'speculative_dspark_block_size': None, 'speculative_dspark_sps_table_path': None, 'speculative_dspark_confidence_sts_path': None, 'speculative_dspark_align_verify_tokens_to_graph_tier': False, 'speculative_accept_threshold_single': 1.0, 'speculative_accept_threshold_acc': 1.0, 'speculative_use_rejection_sampling': False, 'speculative_token_map': None, 'speculative_attention_mode': 'prefill', 'speculative_draft_attention_backend': None, 'speculative_dsa_topk_backend': 'sgl-kernel', 'speculative_draft_kv_cache_dtype': None, 'speculative_draft_window_size': None, 'speculative_moe_runner_backend': 'auto', 'speculative_moe_a2a_backend': None, 'speculative_enable_w4a4_mxfp4_megamoe': None, 'speculative_draft_model_quantization': None, '_speculative_draft_quantization_explicitly_set': False, 'speculative_skip_dp_mlp_sync': False, 'enable_multi_layer_eagle': False, 'speculative_adaptive': False, 'speculative_adaptive_config': None, 'decoupled_spec_bind_endpoint': None, 'decoupled_spec_connect_endpoints': None, 'decoupled_spec_rank': None, 'decoupled_spec_role': 'null', 'spec_trace_dir': None, 'speculative_ngram_min_bfs_breadth': 1, 'speculative_ngram_max_bfs_breadth': 10, 'speculative_ngram_match_type': 'BFS', 'speculative_ngram_max_trie_depth': 18, 'speculative_ngram_capacity': 10000000, 'speculative_ngram_external_corpus_path': None, 'speculative_ngram_external_sam_budget': 0, 'speculative_ngram_external_corpus_max_tokens': 10000000, 'ep_size': 1, 'moe_a2a_backend': 'none', 'enable_w4a4_mxfp4_megamoe': False, 'deepep_v2_mode': 'direct', 'moe_runner_backend': 'auto', 'flashinfer_mxfp4_moe_precision': 'default', 'deepep_mode': 'auto', 'fuseep_mode': 2, 'deepep_dispatcher_output_dtype': 'auto', 'ep_num_redundant_experts': 0, 'ep_dispatch_algorithm': None, 'init_expert_location': 'trivial', 'enable_eplb': False, 'eplb_algorithm': 'auto', 'eplb_rebalance_num_iterations': 1000, 'eplb_rebalance_layers_per_chunk': None, 'eplb_min_rebalancing_utilization_threshold': 1.0, 'expert_distribution_recorder_mode': None, 'expert_distribution_recorder_buffer_size': 1000, 'expert_balancedness_report_mode': 'off', 'deepep_config': None, 'moe_dense_tp_size': None, 'elastic_ep_backend': None, 'enable_elastic_expert_backup': False, 'mooncake_ib_device': None, 'enable_waterfill': False, 'ep_join_mode': None, 'ep_join_rank_offset': 0, 'elastic_ep_initial_size': None, 'max_ep_size': None, 'elastic_ep_scale_timeout': 600, 'disable_flashinfer_cutlass_moe_fp4_allgather': False, 'disable_shared_experts_fusion': False, 'enforce_shared_experts_fusion': False, 'max_mamba_cache_size': None, 'mamba_ssm_dtype': None, 'mamba_max_states_per_path': -1, 'enable_mamba_cache_stochastic_rounding': False, 'mamba_cache_philox_rounds': 0, 'mamba_full_memory_ratio': 0.9, 'mamba_radix_cache_strategy': 'auto', 'uses_mamba_radix_cache': False, 'mamba_track_interval': 256, 'enable_int8_mamba_checkpoint': False, 'int8_mamba_ckpt_size': None, 'linear_attn_backend': 'triton', 'linear_attn_decode_backend': None, 'linear_attn_prefill_backend': None, 'linear_attn_verify_backend': None, 'enable_linear_replayssm': False, 'linear_replayssm_cache_len': 16, 'enable_linear_replayssm_spec': False, 'enable_hierarchical_cache': False, 'hicache_host_memory_mode': 'cache', 'hicache_ratio': 2.0, 'hicache_size': 0, 'hicache_write_policy': 'write_through', 'hicache_io_backend': 'kernel', 'hicache_mem_layout': 'page_first', 'hicache_storage_backend': None, 'hicache_storage_prefetch_policy': 'timeout', 'hicache_storage_backend_extra_config': None, 'hicache_storage_prefetch_retry_poll_interval': 8, 'hicache_storage_prefetch_retry_max_attempts': 8, 'enable_unified_cache_external_linker': False, 'unified_cache_external_linker_backend': 'mooncake', 'enable_hisparse': False, 'hisparse_config': None, 'enable_broadcast_mm_inputs_process': False, 'enable_prefix_mm_cache': False, 'mm_enable_dp_encoder': False, 'mm_process_config': {}, 'mm_processor_worker_num': 0, 'mm_io_worker_num': 0, 'allowed_media_domains': [], 'media_url_max_file_size_mb': 64, 'mm_preprocess_cache_size_mb': None, 'trust_mm_content_hashes': False, 'limit_mm_data_per_request': None, 'enable_mm_global_cache': False, 'image_processor_backend': 'auto', 'mm_global_cache_backend': 'mooncake', 'disable_fast_image_processor': False, 'mm_feature_transport': 'cpu', 'enable_lora': None, 'enable_lora_overlap_loading': None, 'max_lora_rank': None, 'lora_target_modules': None, 'lora_paths': None, 'max_loaded_loras': None, 'max_loras_per_batch': 8, 'lora_eviction_policy': 'lru', 'lora_backend': 'csgmv', 'max_lora_chunk_size': 16, 'experts_shared_outer_loras': None, 'lora_use_virtual_experts': False, 'lora_strict_loading': False, 'lora_drain_wait_threshold': 0.0, 'enable_two_batch_overlap': False, 'enable_single_batch_overlap': False, 'tbo_token_distribution_threshold': 0.48, 'cpu_offload_gb': 0, 'offload_group_size': -1, 'offload_num_in_group': 1, 'offload_prefetch_step': 1, 'offload_mode': 'cpu', 'enable_lmcache': False, 'lmcache_config_file': None, 'enable_flexkv': False, 'flexkv_config_file': None, 'kt_weight_path': None, 'kt_method': 'AMXINT4', 'kt_cpuinfer': None, 'kt_threadpool_count': 2, 'kt_num_gpu_experts': None, 'kt_max_deferred_experts_per_token': None, 'dllm_algorithm': None, 'dllm_algorithm_config': None, 'dllm_fdfo': True, 'disaggregation_mode': 'null', 'disaggregation_transfer_backend': 'mooncake', 'disaggregation_bootstrap_port': 8998, 'disaggregation_ib_device': None, 'disaggregation_decode_enable_radix_cache': False, 'disaggregation_decode_enable_offload_kvcache': False, 'disaggregation_decode_retraction_backup': None, 'num_reserved_decode_tokens': 512, 'disaggregation_decode_extra_slots': None, 'disaggregation_decode_polling_interval': 1, 'optimistic_prefill_attempts': 0, 'encoder_only': False, 'language_only': False, 'language_model_only': False, 'encoder_transfer_backend': 'zmq_to_scheduler', 'encoder_urls': [], 'encoder_bootstrap_port': 8997, 'encoder_register_urls': [], 'enable_adaptive_dispatch_to_encoder': False, 'enable_pdmux': False, 'pdmux_config_path': None, 'sm_group_num': 8, 'startup_weight_load_mode': 'serial', 'custom_weight_loader': [], 'weight_loader_disable_mmap': False, 'weight_loader_prefetch_checkpoints': False, 'weight_loader_prefetch_num_threads': 4, 'weight_loader_drop_cache_after_load': False, 'remote_instance_weight_loader_seed_instance_ip': None, 'remote_instance_weight_loader_seed_instance_service_port': None, 'remote_instance_weight_loader_send_weights_group_ports': None, 'remote_instance_weight_loader_backend': 'nccl', 'remote_instance_weight_loader_start_seed_via_transfer_engine': False, 'engine_info_bootstrap_port': 6789, 'modelexpress_config': None, 'download_dir': None, 'model_checksum': None, 'delete_ckpt_after_loading': False, 'decrypted_config_file': None, 'decrypted_draft_config_file': None, 'checkpoint_engine_wait_weights_before_ready': False, 'enable_prefill_delayer': False, 'prefill_delayer_max_delay_passes': 30, 'prefill_delayer_token_usage_low_watermark': None, 'prefill_delayer_forward_passes_buckets': None, 'prefill_delayer_wait_seconds_buckets': None, 'prefill_delayer_queue_min_ratio': None, 'prefill_delayer_max_delay_ms': None, 'min_free_slots_delay': None, 'enable_deterministic_inference': False, 'rl_on_policy_target': None, 'kv_canary': 'none', 'kv_canary_real_data': 'none', 'kv_canary_sweep_interval': 0, 'enable_dynamic_batch_tokenizer': False, 'dynamic_batch_tokenizer_batch_size': 32, 'dynamic_batch_tokenizer_batch_timeout': 0.002, 'enable_tokenizer_batch_encode': False, 'disable_tokenizer_batch_decode': False, 'debug_tensor_dump_output_folder': None, 'debug_tensor_dump_layers': None, 'debug_tensor_dump_input_file': None, 'enable_memory_saver': False, 'enable_weights_cpu_backup': False, 'enable_draft_weights_cpu_backup': False, 'enable_custom_logit_processor': False, 'enable_return_hidden_states': False, 'return_hidden_states_mode': None, 'enable_return_routed_experts': False, 'enable_return_indexer_topk': False, 'sampling_mask_max_tokens': 4096, 'disable_outlines_disk_cache': False, 'enable_mis': False, 'weight_cache_mode': 'off', 'weight_cache_socket': None, 'weight_cache_timeout': 1800, 'forward_hooks': None, 'msprobe_dump_config': None, 'ple_offload_embedding': None, 'ple_offload_backend': 'pinned', 'ple_offload_dir': None, 'disaggregation_enable_kv_checksum': False, 'enable_pd_role_switch': False, 'speculative_domino_candidate_pool_size': 2048, 'flashinfer_a2a_dispatch_type': None, 'cuda_graph_max_seq_len_prefill': None, 'prefill_kv_cache_dequant_dtype': 'auto', 'dsv4_attn_backend': 'auto', 'otlp_service_name': None, '_radix_eviction_policy_explicitly_set': False, 'hicache_host_memory_fraction': 0.8, 'enable_linker_mla_dedup': False, '_swa_full_tokens_ratio_explicitly_set': False, 'swa_prefix_tails': None, 'enable_encoder_swa_bounded_replay': False, 'enable_decoder_swa_bounded_replay': False, 'enable_response_store': False, 'disaggregation_decode_host_receive_threshold': 1.0, 'qsa_indexer_dtype': 'auto', 'disaggregation_decode_allocation_policy': 'early', 'speculative_boundary_reduction': 'rs+rsv', 'speculative_use_block_verification': False, 'speculative_draft_sink_size': None, 'disable_aiter_allreduce_fusion_in_prefill': False, 'disable_aiter_allreduce_fusion_in_decode': False, 'dsa_triton_union': 0, 'enable_dsa_fused_indexer': None, 'enable_cp_tp_group_sharing': False, 'enable_watermark': False, 'watermark_config': '<redacted>', 'watermark_key': '<redacted>', 'watermark_key_b': '<redacted>', 'watermark_context_window': 4, 'watermark_mixing_probability': 0.5, 'watermark_max_probability': 1.0, 'watermark_default_enabled': False, 'watermark_enforce_all': False, 'grpc_response_timeout_secs': 300}
[2026-10-17 14:56:54] Using default HuggingFace chat template with detected content format: string
[2026-10-17 14:56:59] Init parallel begin.
[2026-10-17 14:56:59] Init parallel ends. elapsed=0.26 s
[2026-10-17 14:56:59] Memory baseline taken. mem usage=0.00 GB
/home/rohit/sglang-lab/.venv/lib/python3.12/site-packages/torch/jit/_script.py:1491: FutureWarning: `torch.jit.script` is deprecated. Please switch to `torch.compile` or `torch.export`.
  warnings.warn(
/home/rohit/sglang-lab/.venv/lib/python3.12/site-packages/torch/jit/_script.py:1643: FutureWarning: `torch.jit.interface` is deprecated. Please use `torch.compile` instead.
  warnings.warn(
/home/rohit/sglang-lab/.venv/lib/python3.12/site-packages/torch/jit/_script.py:1491: FutureWarning: `torch.jit.script` is deprecated. Please switch to `torch.compile` or `torch.export`.
  warnings.warn(
[2026-10-17 14:57:01] Load weight begin. avail mem=113.34 GB
[2026-10-17 14:57:03] Found local HF snapshot for Qwen/Qwen3-30B-A3B-Instruct-2507 at /home/rohit/sglang-lab/hf/hub/models--Qwen--Qwen3-30B-A3B-Instruct-2507/snapshots/0d7cf23991f47feeb3a57ecb4c9cee8ea4a17bfe; skipping download.

Multi-thread loading shards:   0% Completed | 0/16 [00:00<?, ?it/s]
Multi-thread loading shards:   6% Completed | 1/16 [00:22<05:33, 22.21s/it]
Multi-thread loading shards:  12% Completed | 2/16 [00:43<05:01, 21.50s/it]
Multi-thread loading shards:  19% Completed | 3/16 [01:04<04:36, 21.27s/it]
Multi-thread loading shards:  25% Completed | 4/16 [01:25<04:13, 21.13s/it]
Multi-thread loading shards:  31% Completed | 5/16 [01:45<03:50, 20.93s/it]
Multi-thread loading shards:  38% Completed | 6/16 [02:05<03:25, 20.57s/it]
Multi-thread loading shards:  44% Completed | 7/16 [02:25<03:03, 20.37s/it]
Multi-thread loading shards:  50% Completed | 8/16 [02:44<02:38, 19.80s/it]
Multi-thread loading shards:  56% Completed | 9/16 [03:03<02:18, 19.79s/it]
Multi-thread loading shards:  62% Completed | 10/16 [03:23<01:58, 19.81s/it]
Multi-thread loading shards:  69% Completed | 11/16 [03:43<01:39, 19.81s/it]
Multi-thread loading shards:  75% Completed | 12/16 [04:04<01:20, 20.10s/it]
Multi-thread loading shards:  81% Completed | 13/16 [04:24<01:00, 20.23s/it]
Multi-thread loading shards:  88% Completed | 14/16 [04:45<00:40, 20.38s/it]
Multi-thread loading shards:  94% Completed | 15/16 [05:06<00:20, 20.65s/it]
Multi-thread loading shards: 100% Completed | 16/16 [05:13<00:00, 16.32s/it]
Multi-thread loading shards: 100% Completed | 16/16 [05:13<00:00, 19.57s/it]
[2026-10-17 15:02:16] Load weight end. elapsed=315.62 s, type=Qwen3MoeForCausalLM, avail mem=55.84 GB, mem usage=57.50 GB.
[2026-10-17 15:02:17] KV Cache is allocated. dtype: torch.bfloat16, #tokens: 4096, K size: 0.19 GB, V size: 0.19 GB
[2026-10-17 15:02:17] Memory pool end. avail mem=52.74 GB
[2026-10-17 15:02:18] Capture target prefill CUDA graph begin. backend=breakable, num_tokens=[4, 8, 12, 16, 20, 24, 28, 32, 48, 64, 80, 96, 112, 128, 144, 160, 176, 192, 208, 224, 240, 256, 288, 320, 352, 384, 416, 448, 480, 512, 576, 640, 704, 768, 832, 896, 960, 1024, 1280, 1536, 1792, 2048, 2304, 2560, 2816, 3072, 3328, 3584, 3840, 4096], avail mem=52.13 GB

  0%|          | 0/50 [00:00<?, ?it/s]
Capturing prefill shape (num_tokens=4096 avail_mem=52.13 GB):   0%|          | 0/50 [00:00<?, ?it/s][2026-10-17 15:02:20] Using default MoE kernel config. Performance might be sub-optimal! Config file not found at /home/rohit/sglang-lab/.venv/lib/python3.12/site-packages/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_8_0/E=128,N=768,device_name=NVIDIA_GB10.json, you can create them with https://github.com/sgl-project/sglang/tree/main/benchmark/kernels/fused_moe_triton
[2026-10-17 15:02:20] Using default MoE kernel config. Performance might be sub-optimal! Config file not found at /home/rohit/sglang-lab/.venv/lib/python3.12/site-packages/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_8_0/E=128,N=768,device_name=NVIDIA_GB10_down.json, you can create them with https://github.com/sgl-project/sglang/tree/main/benchmark/kernels/fused_moe_triton

Capturing prefill shape (num_tokens=4096 avail_mem=52.13 GB):   2%|▏         | 1/50 [00:02<02:04,  2.53s/it]
Capturing prefill shape (num_tokens=3840 avail_mem=51.17 GB):   2%|▏         | 1/50 [00:02<02:04,  2.53s/it]
Capturing prefill shape (num_tokens=3840 avail_mem=51.17 GB):   4%|▍         | 2/50 [00:03<01:24,  1.76s/it]
Capturing prefill shape (num_tokens=3584 avail_mem=51.02 GB):   4%|▍         | 2/50 [00:03<01:24,  1.76s/it]
Capturing prefill shape (num_tokens=3584 avail_mem=51.02 GB):   6%|▌         | 3/50 [00:04<01:09,  1.47s/it]
Capturing prefill shape (num_tokens=3328 avail_mem=51.00 GB):   6%|▌         | 3/50 [00:04<01:09,  1.47s/it]
Capturing prefill shape (num_tokens=3328 avail_mem=51.00 GB):   8%|▊         | 4/50 [00:05<00:59,  1.30s/it]
Capturing prefill shape (num_tokens=3072 avail_mem=50.99 GB):   8%|▊         | 4/50 [00:05<00:59,  1.30s/it]
Capturing prefill shape (num_tokens=3072 avail_mem=50.99 GB):  10%|█         | 5/50 [00:06<00:53,  1.18s/it]
Capturing prefill shape (num_tokens=2816 avail_mem=50.98 GB):  10%|█         | 5/50 [00:06<00:53,  1.18s/it]
Capturing prefill shape (num_tokens=2816 avail_mem=50.98 GB):  12%|█▏        | 6/50 [00:07<00:47,  1.08s/it]
Capturing prefill shape (num_tokens=2560 avail_mem=50.96 GB):  12%|█▏        | 6/50 [00:07<00:47,  1.08s/it]
Capturing prefill shape (num_tokens=2560 avail_mem=50.96 GB):  14%|█▍        | 7/50 [00:08<00:42,  1.02it/s]
Capturing prefill shape (num_tokens=2304 avail_mem=50.94 GB):  14%|█▍        | 7/50 [00:08<00:42,  1.02it/s]
Capturing prefill shape (num_tokens=2304 avail_mem=50.94 GB):  16%|█▌        | 8/50 [00:09<00:37,  1.12it/s]
Capturing prefill shape (num_tokens=2048 avail_mem=50.93 GB):  16%|█▌        | 8/50 [00:09<00:37,  1.12it/s]
Capturing prefill shape (num_tokens=2048 avail_mem=50.93 GB):  18%|█▊        | 9/50 [00:09<00:32,  1.24it/s]
Capturing prefill shape (num_tokens=1792 avail_mem=50.92 GB):  18%|█▊        | 9/50 [00:09<00:32,  1.24it/s]
Capturing prefill shape (num_tokens=1792 avail_mem=50.92 GB):  20%|██        | 10/50 [00:10<00:28,  1.39it/s]
Capturing prefill shape (num_tokens=1536 avail_mem=50.91 GB):  20%|██        | 10/50 [00:10<00:28,  1.39it/s]
Capturing prefill shape (num_tokens=1536 avail_mem=50.91 GB):  22%|██▏       | 11/50 [00:10<00:24,  1.56it/s]
Capturing prefill shape (num_tokens=1280 avail_mem=50.89 GB):  22%|██▏       | 11/50 [00:10<00:24,  1.56it/s]
Capturing prefill shape (num_tokens=1280 avail_mem=50.89 GB):  24%|██▍       | 12/50 [00:11<00:20,  1.81it/s]
Capturing prefill shape (num_tokens=1024 avail_mem=50.88 GB):  24%|██▍       | 12/50 [00:11<00:20,  1.81it/s]
Capturing prefill shape (num_tokens=1024 avail_mem=50.88 GB):  26%|██▌       | 13/50 [00:11<00:17,  2.09it/s]
Capturing prefill shape (num_tokens=960 avail_mem=50.86 GB):  26%|██▌       | 13/50 [00:11<00:17,  2.09it/s] 
Capturing prefill shape (num_tokens=960 avail_mem=50.86 GB):  28%|██▊       | 14/50 [00:11<00:14,  2.44it/s]
Capturing prefill shape (num_tokens=896 avail_mem=50.85 GB):  28%|██▊       | 14/50 [00:11<00:14,  2.44it/s]
Capturing prefill shape (num_tokens=896 avail_mem=50.85 GB):  30%|███       | 15/50 [00:11<00:12,  2.84it/s]
Capturing prefill shape (num_tokens=832 avail_mem=50.84 GB):  30%|███       | 15/50 [00:11<00:12,  2.84it/s]
Capturing prefill shape (num_tokens=832 avail_mem=50.84 GB):  32%|███▏      | 16/50 [00:12<00:10,  3.20it/s]
Capturing prefill shape (num_tokens=768 avail_mem=50.82 GB):  32%|███▏      | 16/50 [00:12<00:10,  3.20it/s]
Capturing prefill shape (num_tokens=768 avail_mem=50.82 GB):  34%|███▍      | 17/50 [00:12<00:09,  3.60it/s]
Capturing prefill shape (num_tokens=704 avail_mem=50.85 GB):  34%|███▍      | 17/50 [00:12<00:09,  3.60it/s]
Capturing prefill shape (num_tokens=704 avail_mem=50.85 GB):  36%|███▌      | 18/50 [00:12<00:07,  4.02it/s]
Capturing prefill shape (num_tokens=640 avail_mem=50.83 GB):  36%|███▌      | 18/50 [00:12<00:07,  4.02it/s]
Capturing prefill shape (num_tokens=640 avail_mem=50.83 GB):  38%|███▊      | 19/50 [00:12<00:06,  4.50it/s]
Capturing prefill shape (num_tokens=576 avail_mem=50.81 GB):  38%|███▊      | 19/50 [00:12<00:06,  4.50it/s]
Capturing prefill shape (num_tokens=576 avail_mem=50.81 GB):  40%|████      | 20/50 [00:12<00:05,  5.03it/s]
Capturing prefill shape (num_tokens=512 avail_mem=50.80 GB):  40%|████      | 20/50 [00:12<00:05,  5.03it/s]
Capturing prefill shape (num_tokens=512 avail_mem=50.80 GB):  42%|████▏     | 21/50 [00:13<00:05,  5.38it/s]
Capturing prefill shape (num_tokens=480 avail_mem=50.78 GB):  42%|████▏     | 21/50 [00:13<00:05,  5.38it/s]
Capturing prefill shape (num_tokens=480 avail_mem=50.78 GB):  44%|████▍     | 22/50 [00:13<00:04,  5.76it/s]
Capturing prefill shape (num_tokens=448 avail_mem=50.72 GB):  44%|████▍     | 22/50 [00:13<00:04,  5.76it/s]
Capturing prefill shape (num_tokens=448 avail_mem=50.72 GB):  46%|████▌     | 23/50 [00:13<00:04,  6.05it/s]
Capturing prefill shape (num_tokens=416 avail_mem=50.69 GB):  46%|████▌     | 23/50 [00:13<00:04,  6.05it/s]
Capturing prefill shape (num_tokens=416 avail_mem=50.69 GB):  48%|████▊     | 24/50 [00:13<00:04,  6.42it/s]
Capturing prefill shape (num_tokens=384 avail_mem=50.33 GB):  48%|████▊     | 24/50 [00:13<00:04,  6.42it/s]
Capturing prefill shape (num_tokens=384 avail_mem=50.33 GB):  50%|█████     | 25/50 [00:13<00:03,  6.75it/s]
Capturing prefill shape (num_tokens=352 avail_mem=50.31 GB):  50%|█████     | 25/50 [00:13<00:03,  6.75it/s]
Capturing prefill shape (num_tokens=352 avail_mem=50.31 GB):  52%|█████▏    | 26/50 [00:13<00:03,  7.00it/s]
Capturing prefill shape (num_tokens=320 avail_mem=50.62 GB):  52%|█████▏    | 26/50 [00:13<00:03,  7.00it/s]
Capturing prefill shape (num_tokens=320 avail_mem=50.62 GB):  54%|█████▍    | 27/50 [00:13<00:03,  7.04it/s]
Capturing prefill shape (num_tokens=288 avail_mem=50.58 GB):  54%|█████▍    | 27/50 [00:13<00:03,  7.04it/s]
Capturing prefill shape (num_tokens=288 avail_mem=50.58 GB):  56%|█████▌    | 28/50 [00:13<00:03,  7.26it/s]
Capturing prefill shape (num_tokens=256 avail_mem=50.57 GB):  56%|█████▌    | 28/50 [00:13<00:03,  7.26it/s]
Capturing prefill shape (num_tokens=256 avail_mem=50.57 GB):  58%|█████▊    | 29/50 [00:14<00:02,  7.43it/s]
Capturing prefill shape (num_tokens=240 avail_mem=50.61 GB):  58%|█████▊    | 29/50 [00:14<00:02,  7.43it/s]
Capturing prefill shape (num_tokens=240 avail_mem=50.61 GB):  60%|██████    | 30/50 [00:14<00:02,  7.50it/s]
Capturing prefill shape (num_tokens=224 avail_mem=50.63 GB):  60%|██████    | 30/50 [00:14<00:02,  7.50it/s]
Capturing prefill shape (num_tokens=224 avail_mem=50.63 GB):  62%|██████▏   | 31/50 [00:14<00:02,  7.72it/s]
Capturing prefill shape (num_tokens=208 avail_mem=50.62 GB):  62%|██████▏   | 31/50 [00:14<00:02,  7.72it/s]
Capturing prefill shape (num_tokens=208 avail_mem=50.62 GB):  64%|██████▍   | 32/50 [00:14<00:02,  7.81it/s]
Capturing prefill shape (num_tokens=192 avail_mem=50.54 GB):  64%|██████▍   | 32/50 [00:14<00:02,  7.81it/s]
Capturing prefill shape (num_tokens=192 avail_mem=50.54 GB):  66%|██████▌   | 33/50 [00:14<00:02,  7.75it/s]
Capturing prefill shape (num_tokens=176 avail_mem=50.57 GB):  66%|██████▌   | 33/50 [00:14<00:02,  7.75it/s]
Capturing prefill shape (num_tokens=176 avail_mem=50.57 GB):  68%|██████▊   | 34/50 [00:14<00:02,  7.93it/s]
Capturing prefill shape (num_tokens=160 avail_mem=50.58 GB):  68%|██████▊   | 34/50 [00:14<00:02,  7.93it/s]
Capturing prefill shape (num_tokens=160 avail_mem=50.58 GB):  70%|███████   | 35/50 [00:14<00:01,  8.07it/s]
Capturing prefill shape (num_tokens=144 avail_mem=50.13 GB):  70%|███████   | 35/50 [00:14<00:01,  8.07it/s]
Capturing prefill shape (num_tokens=144 avail_mem=50.13 GB):  72%|███████▏  | 36/50 [00:14<00:01,  8.31it/s]
Capturing prefill shape (num_tokens=128 avail_mem=50.12 GB):  72%|███████▏  | 36/50 [00:14<00:01,  8.31it/s]
Capturing prefill shape (num_tokens=128 avail_mem=50.12 GB):  74%|███████▍  | 37/50 [00:15<00:01,  7.98it/s]
Capturing prefill shape (num_tokens=112 avail_mem=50.11 GB):  74%|███████▍  | 37/50 [00:15<00:01,  7.98it/s]
Capturing prefill shape (num_tokens=112 avail_mem=50.11 GB):  76%|███████▌  | 38/50 [00:15<00:01,  8.15it/s]
Capturing prefill shape (num_tokens=96 avail_mem=50.09 GB):  76%|███████▌  | 38/50 [00:15<00:01,  8.15it/s] 
Capturing prefill shape (num_tokens=96 avail_mem=50.09 GB):  78%|███████▊  | 39/50 [00:15<00:01,  8.30it/s]
Capturing prefill shape (num_tokens=80 avail_mem=50.08 GB):  78%|███████▊  | 39/50 [00:15<00:01,  8.30it/s]
Capturing prefill shape (num_tokens=80 avail_mem=50.08 GB):  80%|████████  | 40/50 [00:15<00:01,  8.65it/s]
Capturing prefill shape (num_tokens=64 avail_mem=50.06 GB):  80%|████████  | 40/50 [00:15<00:01,  8.65it/s]
Capturing prefill shape (num_tokens=64 avail_mem=50.06 GB):  82%|████████▏ | 41/50 [00:15<00:01,  8.46it/s]
Capturing prefill shape (num_tokens=48 avail_mem=50.05 GB):  82%|████████▏ | 41/50 [00:15<00:01,  8.46it/s]
Capturing prefill shape (num_tokens=48 avail_mem=50.05 GB):  84%|████████▍ | 42/50 [00:15<00:00,  8.76it/s]
Capturing prefill shape (num_tokens=32 avail_mem=50.03 GB):  84%|████████▍ | 42/50 [00:15<00:00,  8.76it/s]
Capturing prefill shape (num_tokens=32 avail_mem=50.03 GB):  86%|████████▌ | 43/50 [00:15<00:00,  8.64it/s]
Capturing prefill shape (num_tokens=28 avail_mem=50.02 GB):  86%|████████▌ | 43/50 [00:15<00:00,  8.64it/s]
Capturing prefill shape (num_tokens=28 avail_mem=50.02 GB):  88%|████████▊ | 44/50 [00:15<00:00,  8.57it/s]
Capturing prefill shape (num_tokens=24 avail_mem=50.01 GB):  88%|████████▊ | 44/50 [00:15<00:00,  8.57it/s]
Capturing prefill shape (num_tokens=24 avail_mem=50.01 GB):  90%|█████████ | 45/50 [00:16<00:00,  8.92it/s]
Capturing prefill shape (num_tokens=20 avail_mem=49.99 GB):  90%|█████████ | 45/50 [00:16<00:00,  8.92it/s]
Capturing prefill shape (num_tokens=20 avail_mem=49.99 GB):  92%|█████████▏| 46/50 [00:16<00:00,  9.05it/s]
Capturing prefill shape (num_tokens=16 avail_mem=50.42 GB):  92%|█████████▏| 46/50 [00:16<00:00,  9.05it/s]
Capturing prefill shape (num_tokens=16 avail_mem=50.42 GB):  94%|█████████▍| 47/50 [00:16<00:00,  7.90it/s]
Capturing prefill shape (num_tokens=12 avail_mem=50.25 GB):  94%|█████████▍| 47/50 [00:16<00:00,  7.90it/s]
Capturing prefill shape (num_tokens=12 avail_mem=50.25 GB):  96%|█████████▌| 48/50 [00:16<00:00,  8.25it/s]
Capturing prefill shape (num_tokens=8 avail_mem=50.23 GB):  96%|█████████▌| 48/50 [00:16<00:00,  8.25it/s] 
Capturing prefill shape (num_tokens=8 avail_mem=50.23 GB):  98%|█████████▊| 49/50 [00:16<00:00,  8.31it/s]
Capturing prefill shape (num_tokens=4 avail_mem=50.22 GB):  98%|█████████▊| 49/50 [00:16<00:00,  8.31it/s]
Capturing prefill shape (num_tokens=4 avail_mem=50.22 GB): 100%|██████████| 50/50 [00:16<00:00,  8.73it/s]
Capturing prefill shape (num_tokens=4 avail_mem=50.22 GB): 100%|██████████| 50/50 [00:16<00:00,  3.01it/s]
[2026-10-17 15:02:35] Capture target prefill CUDA graph end. elapsed=17.10 s, mem usage=1.38 GB, avail mem=50.74 GB.
[2026-10-17 15:02:35] Capture target decode CUDA graph begin. backend=full, num_tokens_per_req=1, avail mem=50.74 GB

  0%|          | 0/36 [00:00<?, ?it/s]
Capturing batches (bs=256 avail_mem=50.48 GB):   0%|          | 0/36 [00:00<?, ?it/s]
Capturing batches (bs=256 avail_mem=50.48 GB):   3%|▎         | 1/36 [00:00<00:04,  7.28it/s]
Capturing batches (bs=248 avail_mem=50.44 GB):   3%|▎         | 1/36 [00:00<00:04,  7.28it/s]
Capturing batches (bs=248 avail_mem=50.44 GB):   6%|▌         | 2/36 [00:00<00:04,  7.34it/s]
Capturing batches (bs=240 avail_mem=50.42 GB):   6%|▌         | 2/36 [00:00<00:04,  7.34it/s]
Capturing batches (bs=240 avail_mem=50.42 GB):   8%|▊         | 3/36 [00:00<00:04,  7.26it/s]
Capturing batches (bs=232 avail_mem=50.37 GB):   8%|▊         | 3/36 [00:00<00:04,  7.26it/s]
Capturing batches (bs=232 avail_mem=50.37 GB):  11%|█         | 4/36 [00:00<00:04,  7.33it/s]
Capturing batches (bs=224 avail_mem=50.36 GB):  11%|█         | 4/36 [00:00<00:04,  7.33it/s]
Capturing batches (bs=224 avail_mem=50.36 GB):  14%|█▍        | 5/36 [00:00<00:04,  7.37it/s]
Capturing batches (bs=216 avail_mem=50.32 GB):  14%|█▍        | 5/36 [00:00<00:04,  7.37it/s]
Capturing batches (bs=216 avail_mem=50.32 GB):  17%|█▋        | 6/36 [00:00<00:04,  7.45it/s]
Capturing batches (bs=208 avail_mem=50.30 GB):  17%|█▋        | 6/36 [00:00<00:04,  7.45it/s]
Capturing batches (bs=208 avail_mem=50.30 GB):  19%|█▉        | 7/36 [00:00<00:03,  7.58it/s]
Capturing batches (bs=200 avail_mem=50.27 GB):  19%|█▉        | 7/36 [00:00<00:03,  7.58it/s]
Capturing batches (bs=200 avail_mem=50.27 GB):  22%|██▏       | 8/36 [00:01<00:03,  7.67it/s]
Capturing batches (bs=192 avail_mem=50.25 GB):  22%|██▏       | 8/36 [00:01<00:03,  7.67it/s]
Capturing batches (bs=192 avail_mem=50.25 GB):  25%|██▌       | 9/36 [00:01<00:03,  7.74it/s]
Capturing batches (bs=184 avail_mem=50.21 GB):  25%|██▌       | 9/36 [00:01<00:03,  7.74it/s]
Capturing batches (bs=184 avail_mem=50.21 GB):  28%|██▊       | 10/36 [00:01<00:03,  7.80it/s]
Capturing batches (bs=176 avail_mem=50.20 GB):  28%|██▊       | 10/36 [00:01<00:03,  7.80it/s]
Capturing batches (bs=176 avail_mem=50.20 GB):  31%|███       | 11/36 [00:01<00:03,  7.87it/s]
Capturing batches (bs=168 avail_mem=50.16 GB):  31%|███       | 11/36 [00:01<00:03,  7.87it/s]
Capturing batches (bs=168 avail_mem=50.16 GB):  33%|███▎      | 12/36 [00:01<00:03,  7.91it/s]
Capturing batches (bs=160 avail_mem=50.14 GB):  33%|███▎      | 12/36 [00:01<00:03,  7.91it/s]
Capturing batches (bs=160 avail_mem=50.14 GB):  36%|███▌      | 13/36 [00:01<00:02,  8.05it/s]
Capturing batches (bs=152 avail_mem=50.11 GB):  36%|███▌      | 13/36 [00:01<00:02,  8.05it/s]
Capturing batches (bs=152 avail_mem=50.11 GB):  39%|███▉      | 14/36 [00:01<00:02,  8.12it/s]
Capturing batches (bs=144 avail_mem=50.09 GB):  39%|███▉      | 14/36 [00:01<00:02,  8.12it/s]
Capturing batches (bs=144 avail_mem=50.09 GB):  42%|████▏     | 15/36 [00:01<00:02,  8.22it/s]
Capturing batches (bs=136 avail_mem=50.06 GB):  42%|████▏     | 15/36 [00:01<00:02,  8.22it/s]
Capturing batches (bs=136 avail_mem=50.06 GB):  44%|████▍     | 16/36 [00:02<00:02,  8.28it/s]
Capturing batches (bs=128 avail_mem=50.04 GB):  44%|████▍     | 16/36 [00:02<00:02,  8.28it/s]
Capturing batches (bs=128 avail_mem=50.04 GB):  47%|████▋     | 17/36 [00:02<00:02,  8.31it/s]
Capturing batches (bs=120 avail_mem=50.00 GB):  47%|████▋     | 17/36 [00:02<00:02,  8.31it/s]
Capturing batches (bs=120 avail_mem=50.00 GB):  50%|█████     | 18/36 [00:02<00:02,  8.36it/s]
Capturing batches (bs=112 avail_mem=49.98 GB):  50%|█████     | 18/36 [00:02<00:02,  8.36it/s]
Capturing batches (bs=112 avail_mem=49.98 GB):  53%|█████▎    | 19/36 [00:02<00:02,  8.49it/s]
Capturing batches (bs=104 avail_mem=49.95 GB):  53%|█████▎    | 19/36 [00:02<00:02,  8.49it/s]
Capturing batches (bs=104 avail_mem=49.95 GB):  56%|█████▌    | 20/36 [00:02<00:01,  8.59it/s]
Capturing batches (bs=96 avail_mem=49.93 GB):  56%|█████▌    | 20/36 [00:02<00:01,  8.59it/s] 
Capturing batches (bs=96 avail_mem=49.93 GB):  58%|█████▊    | 21/36 [00:02<00:01,  8.76it/s]
Capturing batches (bs=88 avail_mem=49.89 GB):  58%|█████▊    | 21/36 [00:02<00:01,  8.76it/s]
Capturing batches (bs=88 avail_mem=49.89 GB):  61%|██████    | 22/36 [00:02<00:01,  8.84it/s]
Capturing batches (bs=80 avail_mem=49.87 GB):  61%|██████    | 22/36 [00:02<00:01,  8.84it/s]
Capturing batches (bs=80 avail_mem=49.87 GB):  64%|██████▍   | 23/36 [00:02<00:01,  9.03it/s]
Capturing batches (bs=72 avail_mem=49.84 GB):  64%|██████▍   | 23/36 [00:02<00:01,  9.03it/s]
Capturing batches (bs=72 avail_mem=49.84 GB):  67%|██████▋   | 24/36 [00:02<00:01,  9.23it/s]
Capturing batches (bs=64 avail_mem=49.82 GB):  67%|██████▋   | 24/36 [00:02<00:01,  9.23it/s]
Capturing batches (bs=64 avail_mem=49.82 GB):  69%|██████▉   | 25/36 [00:03<00:01,  9.43it/s]
Capturing batches (bs=56 avail_mem=49.78 GB):  69%|██████▉   | 25/36 [00:03<00:01,  9.43it/s]
Capturing batches (bs=56 avail_mem=49.78 GB):  72%|███████▏  | 26/36 [00:03<00:01,  9.58it/s]
Capturing batches (bs=48 avail_mem=49.76 GB):  72%|███████▏  | 26/36 [00:03<00:01,  9.58it/s]
Capturing batches (bs=48 avail_mem=49.76 GB):  75%|███████▌  | 27/36 [00:03<00:00,  9.67it/s]
Capturing batches (bs=40 avail_mem=49.53 GB):  75%|███████▌  | 27/36 [00:03<00:00,  9.67it/s]
Capturing batches (bs=40 avail_mem=49.53 GB):  78%|███████▊  | 28/36 [00:03<00:00,  9.64it/s]
Capturing batches (bs=32 avail_mem=49.73 GB):  78%|███████▊  | 28/36 [00:03<00:00,  9.64it/s]
Capturing batches (bs=24 avail_mem=49.69 GB):  78%|███████▊  | 28/36 [00:03<00:00,  9.64it/s]
Capturing batches (bs=24 avail_mem=49.69 GB):  83%|████████▎ | 30/36 [00:03<00:00,  9.95it/s]
Capturing batches (bs=16 avail_mem=49.67 GB):  83%|████████▎ | 30/36 [00:03<00:00,  9.95it/s]
Capturing batches (bs=12 avail_mem=49.64 GB):  83%|████████▎ | 30/36 [00:03<00:00,  9.95it/s]
Capturing batches (bs=12 avail_mem=49.64 GB):  89%|████████▉ | 32/36 [00:03<00:00, 10.26it/s]
Capturing batches (bs=8 avail_mem=49.62 GB):  89%|████████▉ | 32/36 [00:03<00:00, 10.26it/s] 
Capturing batches (bs=4 avail_mem=49.59 GB):  89%|████████▉ | 32/36 [00:03<00:00, 10.26it/s]
Capturing batches (bs=4 avail_mem=49.59 GB):  94%|█████████▍| 34/36 [00:03<00:00, 10.29it/s]
Capturing batches (bs=2 avail_mem=49.58 GB):  94%|█████████▍| 34/36 [00:03<00:00, 10.29it/s]
Capturing batches (bs=1 avail_mem=49.54 GB):  94%|█████████▍| 34/36 [00:04<00:00, 10.29it/s]
Capturing batches (bs=1 avail_mem=49.54 GB): 100%|██████████| 36/36 [00:04<00:00,  9.61it/s]
Capturing batches (bs=1 avail_mem=49.54 GB): 100%|██████████| 36/36 [00:04<00:00,  8.67it/s]
[2026-10-17 15:02:40] Capture target decode CUDA graph end. bs=[1, 2, 4, 8, 12, 16, 24, 32, 40, 48, 56, 64, 72, 80, 88, 96, 104, 112, 120, 128, 136, 144, 152, 160, 168, 176, 184, 192, 200, 208, 216, 224, 232, 240, 248, 256], elapsed=4.65 s, mem usage=1.34 GB, avail mem=49.41 GB.
[2026-10-17 15:02:40] max_total_num_tokens=4096, chunked_prefill_size=8192, max_prefill_tokens=16384, max_running_requests=2048, context_len=262144, available_gpu_mem=49.41 GB
[2026-10-17 15:02:40] Init Unified Radix Cache. Components: (<ComponentType.FULL: 0>,). Tree Core: RustUnifiedTreeCore
[2026-10-17 15:02:40] Tree cache initialized: source=default impl=UnifiedRadixCache hybrid_swa=False hybrid_ssm=False hicache_attached=False
[2026-10-17 15:02:40] Engine startup timings (s): load_weight=315.62, kv_cache_allocation=0.38, scheduler_e2e=341.35, cuda_graph={prefill=17.10, decode=4.65, target_verify=0.00, draft_prefill=0.00, draft_decode=0.00, draft_extend=0.00}, tokenizer_e2e=346.24
[2026-10-17 15:02:40] INFO:     Started server process [118090]
[2026-10-17 15:02:40] INFO:     Waiting for application startup.
[2026-10-17 15:02:40] Using default chat sampling params from model generation config: {'temperature': 0.7, 'top_k': 20, 'top_p': 0.8}
[2026-10-17 15:02:40] INFO:     Application startup complete.
[2026-10-17 15:02:40] INFO:     Uvicorn running on http://127.0.0.1:30000 (Press CTRL+C to quit)
[2026-10-17 15:02:40] INFO:     127.0.0.1:46288 - "GET /health HTTP/1.1" 503 Service Unavailable
[2026-10-17 15:02:41] INFO:     127.0.0.1:46290 - "GET /model_info HTTP/1.1" 200 OK
[2026-10-17 15:02:41] Prefill batch, #new-seq: 1, #new-token: 6, #cached-token: 0, token usage: 0.00, #running-req: 0, #queue-req: 0, #pending-token: 0, cuda graph: True, input throughput (token/s): 4.04
[2026-10-17 15:02:41] INFO:     127.0.0.1:46296 - "POST /generate HTTP/1.1" 200 OK
[2026-10-17 15:02:41] INFO:     127.0.0.1:46300 - "POST /freeze_gc HTTP/1.1" 200 OK
[2026-10-17 15:02:41] The server is fired up and ready to roll!
[2026-10-17 15:02:42] Prefill batch, #new-seq: 1, #new-token: 1, #cached-token: 0, token usage: 0.00, #running-req: 0, #queue-req: 0, #pending-token: 0, cuda graph: False, input throughput (token/s): 0.74
[2026-10-17 15:02:43] INFO:     127.0.0.1:46312 - "GET /health HTTP/1.1" 200 OK
/home/rohit/sglang-lab/.venv/lib/python3.12/site-packages/sglang/srt/entrypoints/http_server.py:476: FastAPIDeprecationWarning: ORJSONResponse is deprecated, FastAPI now serializes data directly to JSON bytes via Pydantic when a return type or response model is set, which is faster and doesn't need a custom response class. Read more in the FastAPI docs: https://fastapi.tiangolo.com/advanced/custom-response/#orjson-or-response-model and https://fastapi.tiangolo.com/tutorial/response-model/
  return await original_handler(ORJSONRequest(request.scope, request.receive))
[2026-10-17 15:02:50] INFO:     127.0.0.1:49822 - "GET /v1/models HTTP/1.1" 200 OK
[2026-10-17 15:02:50] INFO:     127.0.0.1:49838 - "GET /v1/models HTTP/1.1" 200 OK
[2026-10-17 15:02:50] INFO:     127.0.0.1:49848 - "GET /model_info HTTP/1.1" 200 OK
[2026-10-17 15:02:55] INFO:     127.0.0.1:49862 - "POST /generate HTTP/1.1" 200 OK
[2026-10-17 15:02:55] Prefill batch, #new-seq: 1, #new-token: 256, #cached-token: 0, token usage: 0.06, #running-req: 0, #queue-req: 0, #pending-token: 0, cuda graph: True, input throughput (token/s): 20.48
[2026-10-17 15:02:57] INFO:     127.0.0.1:49868 - "POST /generate HTTP/1.1" 200 OK
[2026-10-17 15:02:57] INFO:     127.0.0.1:49870 - "POST /generate HTTP/1.1" 200 OK
[2026-10-17 15:02:57] INFO:     127.0.0.1:49880 - "POST /generate HTTP/1.1" 200 OK
[2026-10-17 15:02:57] INFO:     127.0.0.1:49884 - "POST /generate HTTP/1.1" 200 OK
[2026-10-17 15:02:57] INFO:     127.0.0.1:49890 - "POST /generate HTTP/1.1" 200 OK
[2026-10-17 15:02:57] INFO:     127.0.0.1:49898 - "POST /generate HTTP/1.1" 200 OK
[2026-10-17 15:02:57] INFO:     127.0.0.1:49908 - "POST /generate HTTP/1.1" 200 OK
[2026-10-17 15:02:57] INFO:     127.0.0.1:49920 - "POST /generate HTTP/1.1" 200 OK
[2026-10-17 15:02:57] INFO:     127.0.0.1:49936 - "POST /generate HTTP/1.1" 200 OK
[2026-10-17 15:02:57] INFO:     127.0.0.1:49946 - "POST /generate HTTP/1.1" 200 OK
[2026-10-17 15:02:57] INFO:     127.0.0.1:49956 - "POST /generate HTTP/1.1" 200 OK
[2026-10-17 15:02:57] INFO:     127.0.0.1:49962 - "POST /generate HTTP/1.1" 200 OK
[2026-10-17 15:02:57] INFO:     127.0.0.1:49966 - "POST /generate HTTP/1.1" 200 OK
[2026-10-17 15:02:57] INFO:     127.0.0.1:49972 - "POST /generate HTTP/1.1" 200 OK
[2026-10-17 15:02:57] INFO:     127.0.0.1:49978 - "POST /generate HTTP/1.1" 200 OK
[2026-10-17 15:02:57] INFO:     127.0.0.1:49984 - "POST /generate HTTP/1.1" 200 OK
[2026-10-17 15:02:57] Prefill batch, #new-seq: 1, #new-token: 1, #cached-token: 255, token usage: 0.28, #running-req: 0, #queue-req: 11, #pending-token: 0, cuda graph: False, input throughput (token/s): 0.49
[2026-10-17 15:02:57] Prefill batch, #new-seq: 4, #new-token: 896, #cached-token: 0, token usage: 0.28, #running-req: 1, #queue-req: 11, #pending-token: 2808, cuda graph: True, input throughput (token/s): 2927.72
[2026-10-17 15:02:58] Decode batch, #running-req: 5, #token: 1167, token usage: 0.28, cuda graph: True, gen throughput (token/s): 2.69, #queue-req: 11
[2026-10-17 15:03:01] Decode batch, #running-req: 5, #token: 1367, token usage: 0.33, cuda graph: True, gen throughput (token/s): 62.02, #queue-req: 11
[2026-10-17 15:03:04] Decode batch, #running-req: 5, #token: 1567, token usage: 0.38, cuda graph: True, gen throughput (token/s): 62.79, #queue-req: 11
[2026-10-17 15:03:07] Decode batch, #running-req: 5, #token: 1767, token usage: 0.43, cuda graph: True, gen throughput (token/s): 61.96, #queue-req: 11
[2026-10-17 15:03:10] Decode batch, #running-req: 5, #token: 1967, token usage: 0.48, cuda graph: True, gen throughput (token/s): 61.47, #queue-req: 11
[2026-10-17 15:03:14] Decode batch, #running-req: 5, #token: 2167, token usage: 0.53, cuda graph: True, gen throughput (token/s): 62.03, #queue-req: 11
[2026-10-17 15:03:17] Decode batch, #running-req: 5, #token: 2367, token usage: 0.58, cuda graph: True, gen throughput (token/s): 62.81, #queue-req: 11
[2026-10-17 15:03:20] Decode batch, #running-req: 5, #token: 2567, token usage: 0.63, cuda graph: True, gen throughput (token/s): 62.43, #queue-req: 11
[2026-10-17 15:03:23] Decode batch, #running-req: 5, #token: 2767, token usage: 0.68, cuda graph: True, gen throughput (token/s): 63.00, #queue-req: 11
[2026-10-17 15:03:27] Decode batch, #running-req: 5, #token: 2967, token usage: 0.72, cuda graph: True, gen throughput (token/s): 59.71, #queue-req: 11
[2026-10-17 15:03:30] Decode batch, #running-req: 5, #token: 3167, token usage: 0.77, cuda graph: True, gen throughput (token/s): 57.98, #queue-req: 11
[2026-10-17 15:03:33] Decode batch, #running-req: 5, #token: 3367, token usage: 0.82, cuda graph: True, gen throughput (token/s): 62.97, #queue-req: 11
[2026-10-17 15:03:36] Decode batch, #running-req: 5, #token: 3567, token usage: 0.87, cuda graph: True, gen throughput (token/s): 62.66, #queue-req: 11
[2026-10-17 15:03:39] INFO:     127.0.0.1:41412 - "POST /generate HTTP/1.1" 200 OK
[2026-10-17 15:03:39] INFO:     127.0.0.1:41428 - "POST /generate HTTP/1.1" 200 OK
[2026-10-17 15:03:39] INFO:     127.0.0.1:41442 - "POST /generate HTTP/1.1" 200 OK
[2026-10-17 15:03:39] INFO:     127.0.0.1:41454 - "POST /generate HTTP/1.1" 200 OK
[2026-10-17 15:03:39] INFO:     127.0.0.1:41462 - "POST /generate HTTP/1.1" 200 OK
[2026-10-17 15:03:39] Prefill batch, #new-seq: 5, #new-token: 1271, #cached-token: 2, token usage: 0.31, #running-req: 0, #queue-req: 7, #pending-token: 1535, cuda graph: True, input throughput (token/s): 30.44
[2026-10-17 15:03:40] Decode batch, #running-req: 5, #token: 1332, token usage: 0.33, cuda graph: True, gen throughput (token/s): 54.01, #queue-req: 11
[2026-10-17 15:03:44] Decode batch, #running-req: 5, #token: 1532, token usage: 0.37, cuda graph: True, gen throughput (token/s): 56.40, #queue-req: 11
[2026-10-17 15:03:47] Decode batch, #running-req: 5, #token: 1732, token usage: 0.42, cuda graph: True, gen throughput (token/s): 57.06, #queue-req: 11
[2026-10-17 15:03:51] Decode batch, #running-req: 5, #token: 1932, token usage: 0.47, cuda graph: True, gen throughput (token/s): 57.56, #queue-req: 11
[2026-10-17 15:03:54] Decode batch, #running-req: 5, #token: 2132, token usage: 0.52, cuda graph: True, gen throughput (token/s): 57.00, #queue-req: 11
[2026-10-17 15:03:58] Decode batch, #running-req: 5, #token: 2332, token usage: 0.57, cuda graph: True, gen throughput (token/s): 57.37, #queue-req: 11
[2026-10-17 15:04:01] Decode batch, #running-req: 5, #token: 2532, token usage: 0.62, cuda graph: True, gen throughput (token/s): 55.77, #queue-req: 11
[2026-10-17 15:04:05] Decode batch, #running-req: 5, #token: 2732, token usage: 0.67, cuda graph: True, gen throughput (token/s): 55.99, #queue-req: 11
[2026-10-17 15:04:08] Decode batch, #running-req: 5, #token: 2932, token usage: 0.72, cuda graph: True, gen throughput (token/s): 56.15, #queue-req: 11
[2026-10-17 15:04:12] Decode batch, #running-req: 5, #token: 3132, token usage: 0.76, cuda graph: True, gen throughput (token/s): 56.38, #queue-req: 11
[2026-10-17 15:04:15] Decode batch, #running-req: 5, #token: 3332, token usage: 0.81, cuda graph: True, gen throughput (token/s): 56.42, #queue-req: 11
[2026-10-17 15:04:19] Decode batch, #running-req: 5, #token: 3532, token usage: 0.86, cuda graph: True, gen throughput (token/s): 56.12, #queue-req: 11
[2026-10-17 15:04:23] Decode batch, #running-req: 5, #token: 3732, token usage: 0.91, cuda graph: True, gen throughput (token/s): 55.50, #queue-req: 11
[2026-10-17 15:04:24] INFO:     127.0.0.1:50480 - "POST /generate HTTP/1.1" 200 OK
[2026-10-17 15:04:24] INFO:     127.0.0.1:50482 - "POST /generate HTTP/1.1" 200 OK
[2026-10-17 15:04:24] INFO:     127.0.0.1:50484 - "POST /generate HTTP/1.1" 200 OK
[2026-10-17 15:04:24] INFO:     127.0.0.1:50498 - "POST /generate HTTP/1.1" 200 OK
[2026-10-17 15:04:24] INFO:     127.0.0.1:50502 - "POST /generate HTTP/1.1" 200 OK
[2026-10-17 15:04:25] Prefill batch, #new-seq: 5, #new-token: 1279, #cached-token: 0, token usage: 0.31, #running-req: 0, #queue-req: 6, #pending-token: 1522, cuda graph: True, input throughput (token/s): 28.03
[2026-10-17 15:04:26] Decode batch, #running-req: 5, #token: 1384, token usage: 0.34, cuda graph: True, gen throughput (token/s): 51.36, #queue-req: 11
[2026-10-17 15:04:30] Decode batch, #running-req: 5, #token: 1584, token usage: 0.39, cuda graph: True, gen throughput (token/s): 58.37, #queue-req: 11
[2026-10-17 15:04:33] Decode batch, #running-req: 5, #token: 1784, token usage: 0.44, cuda graph: True, gen throughput (token/s): 57.15, #queue-req: 11
[2026-10-17 15:04:37] Decode batch, #running-req: 5, #token: 1984, token usage: 0.48, cuda graph: True, gen throughput (token/s): 57.63, #queue-req: 11
[2026-10-17 15:04:40] Decode batch, #running-req: 5, #token: 2184, token usage: 0.53, cuda graph: True, gen throughput (token/s): 57.85, #queue-req: 11
[2026-10-17 15:04:44] Decode batch, #running-req: 5, #token: 2384, token usage: 0.58, cuda graph: True, gen throughput (token/s): 58.43, #queue-req: 11
[2026-10-17 15:04:47] Decode batch, #running-req: 5, #token: 2584, token usage: 0.63, cuda graph: True, gen throughput (token/s): 58.12, #queue-req: 11
[2026-10-17 15:04:51] Decode batch, #running-req: 5, #token: 2784, token usage: 0.68, cuda graph: True, gen throughput (token/s): 58.72, #queue-req: 11
[2026-10-17 15:04:54] Decode batch, #running-req: 5, #token: 2984, token usage: 0.73, cuda graph: True, gen throughput (token/s): 59.49, #queue-req: 11
[2026-10-17 15:04:57] Decode batch, #running-req: 5, #token: 3184, token usage: 0.78, cuda graph: True, gen throughput (token/s): 58.37, #queue-req: 11
[2026-10-17 15:05:01] Decode batch, #running-req: 5, #token: 3384, token usage: 0.83, cuda graph: True, gen throughput (token/s): 57.32, #queue-req: 11
[2026-10-17 15:05:04] Decode batch, #running-req: 5, #token: 3584, token usage: 0.88, cuda graph: True, gen throughput (token/s): 58.64, #queue-req: 11
[2026-10-17 15:05:08] Decode batch, #running-req: 5, #token: 3784, token usage: 0.92, cuda graph: True, gen throughput (token/s): 59.28, #queue-req: 11
[2026-10-17 15:05:09] INFO:     127.0.0.1:39380 - "POST /generate HTTP/1.1" 200 OK
[2026-10-17 15:05:09] INFO:     127.0.0.1:39394 - "POST /generate HTTP/1.1" 200 OK
[2026-10-17 15:05:09] INFO:     127.0.0.1:39396 - "POST /generate HTTP/1.1" 200 OK
[2026-10-17 15:05:09] INFO:     127.0.0.1:39400 - "POST /generate HTTP/1.1" 200 OK
[2026-10-17 15:05:09] INFO:     127.0.0.1:39408 - "POST /generate HTTP/1.1" 200 OK
[2026-10-17 15:05:09] Prefill batch, #new-seq: 5, #new-token: 1277, #cached-token: 1, token usage: 0.31, #running-req: 0, #queue-req: 6, #pending-token: 1524, cuda graph: True, input throughput (token/s): 28.85
[2026-10-17 15:05:12] Decode batch, #running-req: 5, #token: 1428, token usage: 0.35, cuda graph: True, gen throughput (token/s): 50.63, #queue-req: 11
[2026-10-17 15:05:15] Decode batch, #running-req: 5, #token: 1628, token usage: 0.40, cuda graph: True, gen throughput (token/s): 54.71, #queue-req: 11
[2026-10-17 15:05:19] Decode batch, #running-req: 5, #token: 1828, token usage: 0.45, cuda graph: True, gen throughput (token/s): 53.69, #queue-req: 11
[2026-10-17 15:05:23] Decode batch, #running-req: 5, #token: 2028, token usage: 0.50, cuda graph: True, gen throughput (token/s): 53.16, #queue-req: 11
[2026-10-17 15:05:26] Decode batch, #running-req: 5, #token: 2228, token usage: 0.54, cuda graph: True, gen throughput (token/s): 53.73, #queue-req: 11
[2026-10-17 15:05:30] Decode batch, #running-req: 5, #token: 2428, token usage: 0.59, cuda graph: True, gen throughput (token/s): 55.82, #queue-req: 11
[2026-10-17 15:05:34] Decode batch, #running-req: 5, #token: 2628, token usage: 0.64, cuda graph: True, gen throughput (token/s): 53.37, #queue-req: 11
[2026-10-17 15:05:38] Decode batch, #running-req: 5, #token: 2828, token usage: 0.69, cuda graph: True, gen throughput (token/s): 53.32, #queue-req: 11
[2026-10-17 15:05:41] Decode batch, #running-req: 5, #token: 3028, token usage: 0.74, cuda graph: True, gen throughput (token/s): 55.35, #queue-req: 11
[2026-10-17 15:05:45] Decode batch, #running-req: 5, #token: 3228, token usage: 0.79, cuda graph: True, gen throughput (token/s): 54.32, #queue-req: 11
[2026-10-17 15:05:48] Decode batch, #running-req: 5, #token: 3428, token usage: 0.84, cuda graph: True, gen throughput (token/s): 54.59, #queue-req: 11
[2026-10-17 15:05:52] Decode batch, #running-req: 5, #token: 3628, token usage: 0.89, cuda graph: True, gen throughput (token/s): 54.55, #queue-req: 11
[2026-10-17 15:05:56] Decode batch, #running-req: 5, #token: 3828, token usage: 0.93, cuda graph: True, gen throughput (token/s): 54.27, #queue-req: 11
[2026-10-17 15:05:56] INFO:     127.0.0.1:58892 - "POST /generate HTTP/1.1" 200 OK
[2026-10-17 15:05:56] Prefill batch, #new-seq: 5, #new-token: 1267, #cached-token: 1, token usage: 0.31, #running-req: 0, #queue-req: 6, #pending-token: 1536, cuda graph: True, input throughput (token/s): 26.69
[2026-10-17 15:06:00] Decode batch, #running-req: 5, #token: 1463, token usage: 0.36, cuda graph: True, gen throughput (token/s): 52.20, #queue-req: 7
[2026-10-17 15:06:03] Decode batch, #running-req: 5, #token: 1663, token usage: 0.41, cuda graph: True, gen throughput (token/s): 59.57, #queue-req: 7
[2026-10-17 15:06:06] Decode batch, #running-req: 5, #token: 1863, token usage: 0.45, cuda graph: True, gen throughput (token/s): 57.88, #queue-req: 7
[2026-10-17 15:06:10] Decode batch, #running-req: 5, #token: 2063, token usage: 0.50, cuda graph: True, gen throughput (token/s): 56.78, #queue-req: 7
[2026-10-17 15:06:14] Decode batch, #running-req: 5, #token: 2263, token usage: 0.55, cuda graph: True, gen throughput (token/s): 56.44, #queue-req: 7
[2026-10-17 15:06:17] Decode batch, #running-req: 5, #token: 2463, token usage: 0.60, cuda graph: True, gen throughput (token/s): 56.15, #queue-req: 7
[2026-10-17 15:06:21] Decode batch, #running-req: 5, #token: 2663, token usage: 0.65, cuda graph: True, gen throughput (token/s): 56.40, #queue-req: 7
[2026-10-17 15:06:24] Decode batch, #running-req: 5, #token: 2863, token usage: 0.70, cuda graph: True, gen throughput (token/s): 56.68, #queue-req: 7
[2026-10-17 15:06:28] Decode batch, #running-req: 5, #token: 3063, token usage: 0.75, cuda graph: True, gen throughput (token/s): 56.67, #queue-req: 7
[2026-10-17 15:06:31] Decode batch, #running-req: 5, #token: 3263, token usage: 0.80, cuda graph: True, gen throughput (token/s): 56.43, #queue-req: 7
[2026-10-17 15:06:35] Decode batch, #running-req: 5, #token: 3463, token usage: 0.85, cuda graph: True, gen throughput (token/s): 56.23, #queue-req: 7
[2026-10-17 15:06:38] Decode batch, #running-req: 5, #token: 3663, token usage: 0.89, cuda graph: True, gen throughput (token/s): 56.95, #queue-req: 7
[2026-10-17 15:06:42] Prefill batch, #new-seq: 5, #new-token: 1278, #cached-token: 2, token usage: 0.31, #running-req: 0, #queue-req: 2, #pending-token: 511, cuda graph: True, input throughput (token/s): 28.23
[2026-10-17 15:06:42] Decode batch, #running-req: 5, #token: 1320, token usage: 0.32, cuda graph: True, gen throughput (token/s): 49.24, #queue-req: 2
[2026-10-17 15:06:46] Decode batch, #running-req: 5, #token: 1520, token usage: 0.37, cuda graph: True, gen throughput (token/s): 52.25, #queue-req: 2
[2026-10-17 15:06:50] Decode batch, #running-req: 5, #token: 1720, token usage: 0.42, cuda graph: True, gen throughput (token/s): 52.53, #queue-req: 2
[2026-10-17 15:06:54] Decode batch, #running-req: 5, #token: 1920, token usage: 0.47, cuda graph: True, gen throughput (token/s): 51.87, #queue-req: 2
[2026-10-17 15:06:58] Decode batch, #running-req: 5, #token: 2120, token usage: 0.52, cuda graph: True, gen throughput (token/s): 52.69, #queue-req: 2
[2026-10-17 15:07:01] Decode batch, #running-req: 5, #token: 2320, token usage: 0.57, cuda graph: True, gen throughput (token/s): 53.65, #queue-req: 2
[2026-10-17 15:07:05] Decode batch, #running-req: 5, #token: 2520, token usage: 0.62, cuda graph: True, gen throughput (token/s): 52.62, #queue-req: 2
[2026-10-17 15:07:09] Decode batch, #running-req: 5, #token: 2720, token usage: 0.66, cuda graph: True, gen throughput (token/s): 52.15, #queue-req: 2
[2026-10-17 15:07:13] Decode batch, #running-req: 5, #token: 2920, token usage: 0.71, cuda graph: True, gen throughput (token/s): 52.68, #queue-req: 2
[2026-10-17 15:07:17] Decode batch, #running-req: 5, #token: 3120, token usage: 0.76, cuda graph: True, gen throughput (token/s): 51.72, #queue-req: 2
[2026-10-17 15:07:21] Decode batch, #running-req: 5, #token: 3320, token usage: 0.81, cuda graph: True, gen throughput (token/s): 52.27, #queue-req: 2
[2026-10-17 15:07:24] Decode batch, #running-req: 5, #token: 3520, token usage: 0.86, cuda graph: True, gen throughput (token/s): 52.17, #queue-req: 2
[2026-10-17 15:07:28] Decode batch, #running-req: 5, #token: 3720, token usage: 0.91, cuda graph: True, gen throughput (token/s): 51.97, #queue-req: 2
[2026-10-17 15:07:31] Prefill batch, #new-seq: 2, #new-token: 511, #cached-token: 0, token usage: 0.13, #running-req: 0, #queue-req: 0, #pending-token: 0, cuda graph: True, input throughput (token/s): 10.40
[2026-10-17 15:07:32] Decode batch, #running-req: 2, #token: 545, token usage: 0.13, cuda graph: True, gen throughput (token/s): 45.51, #queue-req: 0
[2026-10-17 15:07:33] Decode batch, #running-req: 2, #token: 625, token usage: 0.15, cuda graph: True, gen throughput (token/s): 44.43, #queue-req: 0
[2026-10-17 15:07:35] Decode batch, #running-req: 2, #token: 705, token usage: 0.17, cuda graph: True, gen throughput (token/s): 43.87, #queue-req: 0
[2026-10-17 15:07:37] Decode batch, #running-req: 2, #token: 785, token usage: 0.19, cuda graph: True, gen throughput (token/s): 43.61, #queue-req: 0
[2026-10-17 15:07:39] Decode batch, #running-req: 2, #token: 865, token usage: 0.21, cuda graph: True, gen throughput (token/s): 43.47, #queue-req: 0
[2026-10-17 15:07:41] Decode batch, #running-req: 2, #token: 945, token usage: 0.23, cuda graph: True, gen throughput (token/s): 43.72, #queue-req: 0
[2026-10-17 15:07:43] Decode batch, #running-req: 2, #token: 1025, token usage: 0.25, cuda graph: True, gen throughput (token/s): 43.71, #queue-req: 0
[2026-10-17 15:07:44] Decode batch, #running-req: 2, #token: 1105, token usage: 0.27, cuda graph: True, gen throughput (token/s): 44.40, #queue-req: 0
[2026-10-17 15:07:46] Decode batch, #running-req: 2, #token: 1185, token usage: 0.29, cuda graph: True, gen throughput (token/s): 43.99, #queue-req: 0
[2026-10-17 15:07:48] Decode batch, #running-req: 2, #token: 1265, token usage: 0.31, cuda graph: True, gen throughput (token/s): 43.98, #queue-req: 0
[2026-10-17 15:07:50] Decode batch, #running-req: 2, #token: 1345, token usage: 0.33, cuda graph: True, gen throughput (token/s): 43.87, #queue-req: 0
[2026-10-17 15:07:52] Decode batch, #running-req: 2, #token: 1425, token usage: 0.35, cuda graph: True, gen throughput (token/s): 43.44, #queue-req: 0
[2026-10-17 15:07:53] Decode batch, #running-req: 2, #token: 1505, token usage: 0.37, cuda graph: True, gen throughput (token/s): 43.38, #queue-req: 0
[2026-10-17 15:07:54] INFO:     127.0.0.1:53246 - "GET /server_info HTTP/1.1" 200 OK
[2026-10-17 15:07:54] INFO:     127.0.0.1:53256 - "GET /server_info HTTP/1.1" 200 OK
[2026-10-17 15:07:56] INFO:     127.0.0.1:53272 - "GET /metrics HTTP/1.1" 200 OK
[2026-10-17 15:07:56] SIGTERM received. signum=None frame=None. Draining requests and shutting down...
[2026-10-17 15:07:56] Gracefully exiting... Remaining number of requests 0. Remaining requests remaining_rids=[].
[2026-10-17 15:07:56] kill_process_tree called: parent_pid=118090, include_parent=False, pid=118090
[2026-10-17 15:07:56] INFO:     Shutting down
[2026-10-17 15:07:56] INFO:     Waiting for application shutdown.
[2026-10-17 15:07:56] INFO:     Application shutdown complete.
[2026-10-17 15:07:56] INFO:     Finished server process [118090]
[2026-10-17 15:07:56] kill_process_tree called: parent_pid=118090, include_parent=False, pid=118090