Home Verkennen Help
Registreren Inloggen
CRBC-MaaS-Platform-Project
/
LQDeployConfig
4
0
Vork 0
Bestanden Issues 0 Pull-aanvragen 0 Wiki
Aftakking: master
Aftakkingen Labels
LQDeployConfig
/
prod
/
models
/
sglang
/
logs
/
qwen3-35b-server.log

qwen3-35b-server.log 22 KB
Permalink Geschiedenis Ruwe

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869 
  1. [2026-03-27 16:29:03] WARNING server_args.py:1748: Disabling overlap schedule since mamba no_buffer is not compatible with overlap schedule, try to use --disable-radix-cache if overlap schedule is necessary
    2. 

  2. [2026-03-27 16:29:03] INFO server_args.py:1835: Attention backend not specified. Use fa3 backend by default.
    3. 

  3. [2026-03-27 16:29:04] server_args=ServerArgs(model_path='/model/Qwen3.5-35B-A3B', tokenizer_path='/model/Qwen3.5-35B-A3B', tokenizer_mode='auto', tokenizer_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', host='0.0.0.0', port=30000, fastapi_root_path='', grpc_mode=False, skip_server_warmup=False, warmups=None, nccl_port=None, checkpoint_engine_wait_weights_before_ready=False, 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, mem_fraction_static=0.8387624218749999, max_running_requests=None, max_queued_requests=None, max_total_tokens=None, chunked_prefill_size=8192, enable_dynamic_chunking=False, max_prefill_tokens=16384, prefill_max_requests=None, schedule_policy='fcfs', enable_priority_scheduling=False, abort_on_priority_when_disabled=False, schedule_low_priority_values_first=False, priority_scheduling_preemption_threshold=10, schedule_conservativeness=1.0, page_size=1, swa_full_tokens_ratio=0.8, disable_hybrid_swa_memory=False, radix_eviction_policy='lru', 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, device='cuda', tp_size=1, pp_size=1, pp_max_micro_batch_size=None, pp_async_batch_depth=0, stream_interval=1, stream_output=False, random_seed=694992898, constrained_json_whitespace_pattern=None, constrained_json_disable_any_whitespace=False, watchdog_timeout=300, soft_watchdog_timeout=None, dist_timeout=None, download_dir=None, model_checksum=None, base_gpu_id=0, gpu_id_step=1, sleep_on_idle=False, custom_sigquit_handler=None, 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=False, enable_metrics_for_all_schedulers=False, 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, collect_tokens_histogram=False, 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, enable_trace=False, otlp_traces_endpoint='localhost:4317', export_metrics_to_file=False, export_metrics_to_file_dir=None, api_key='lq123456', admin_api_key=None, served_model_name='/model/Qwen3.5-35B-A3B', weight_version='default', chat_template=None, hf_chat_template_name=None, completion_template=None, file_storage_path='sglang_storage', enable_cache_report=False, reasoning_parser=None, tool_call_parser=None, tool_server=None, sampling_defaults='model', dp_size=1, load_balance_method='round_robin', attn_cp_size=1, moe_dp_size=1, dist_init_addr=None, nnodes=1, node_rank=0, json_model_override_args='{}', preferred_sampling_params=None, 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, attention_backend='fa3', decode_attention_backend=None, prefill_attention_backend=None, sampling_backend='flashinfer', grammar_backend='xgrammar', mm_attention_backend=None, fp8_gemm_runner_backend='auto', fp4_gemm_runner_backend='flashinfer_cutlass', nsa_prefill_backend=None, nsa_decode_backend=None, disable_flashinfer_autotune=False, mamba_backend='triton', speculative_algorithm=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_accept_threshold_single=1.0, speculative_accept_threshold_acc=1.0, speculative_token_map=None, speculative_attention_mode='prefill', speculative_draft_attention_backend=None, speculative_moe_runner_backend='auto', speculative_moe_a2a_backend=None, speculative_draft_model_quantization=None, speculative_ngram_min_match_window_size=1, speculative_ngram_max_match_window_size=12, speculative_ngram_min_bfs_breadth=1, speculative_ngram_max_bfs_breadth=10, speculative_ngram_match_type='BFS', speculative_ngram_branch_length=18, speculative_ngram_capacity=10000000, enable_multi_layer_eagle=False, ep_size=1, moe_a2a_backend='none', moe_runner_backend='auto', flashinfer_mxfp4_moe_precision='default', enable_flashinfer_allreduce_fusion=False, deepep_mode='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, enable_expert_distribution_metrics=False, deepep_config=None, moe_dense_tp_size=None, elastic_ep_backend=None, mooncake_ib_device=None, max_mamba_cache_size=None, mamba_ssm_dtype=None, mamba_full_memory_ratio=0.9, mamba_scheduler_strategy='no_buffer', mamba_track_interval=256, enable_hierarchical_cache=False, hicache_ratio=2.0, hicache_size=0, hicache_write_policy='write_through', hicache_io_backend='kernel', hicache_mem_layout='layer_first', disable_hicache_numa_detect=False, hicache_storage_backend=None, hicache_storage_prefetch_policy='best_effort', hicache_storage_backend_extra_config=None, hierarchical_sparse_attention_extra_config=None, enable_lmcache=False, 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, enable_double_sparsity=False, ds_channel_config_path=None, ds_heavy_channel_num=32, ds_heavy_token_num=256, ds_heavy_channel_type='qk', ds_sparse_decode_threshold=4096, cpu_offload_gb=0, offload_group_size=-1, offload_num_in_group=1, offload_prefetch_step=1, offload_mode='cpu', multi_item_scoring_delimiter=None, disable_radix_cache=False, cuda_graph_max_bs=256, cuda_graph_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], disable_cuda_graph=False, disable_cuda_graph_padding=False, enable_profile_cuda_graph=False, enable_cudagraph_gc=False, enable_layerwise_nvtx_marker=False, enable_nccl_nvls=False, enable_symm_mem=False, disable_flashinfer_cutlass_moe_fp4_allgather=False, enable_tokenizer_batch_encode=False, disable_tokenizer_batch_decode=False, disable_outlines_disk_cache=False, disable_custom_all_reduce=False, enable_mscclpp=False, enable_torch_symm_mem=False, disable_overlap_schedule=True, enable_mixed_chunk=False, enable_dp_attention=False, enable_dp_lm_head=False, enable_two_batch_overlap=False, enable_single_batch_overlap=False, tbo_token_distribution_threshold=0.48, enable_torch_compile=False, enable_piecewise_cuda_graph=False, enable_torch_compile_debug_mode=False, torch_compile_max_bs=32, piecewise_cuda_graph_max_tokens=8192, piecewise_cuda_graph_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, 4608, 5120, 5632, 6144, 6656, 7168, 7680, 8192], piecewise_cuda_graph_compiler='eager', torchao_config='', enable_nan_detection=False, enable_p2p_check=False, triton_attention_reduce_in_fp32=False, triton_attention_num_kv_splits=8, triton_attention_split_tile_size=None, num_continuous_decode_steps=1, delete_ckpt_after_loading=False, enable_memory_saver=False, enable_weights_cpu_backup=False, enable_draft_weights_cpu_backup=False, allow_auto_truncate=False, enable_custom_logit_processor=False, flashinfer_mla_disable_ragged=False, disable_shared_experts_fusion=False, disable_chunked_prefix_cache=False, disable_fast_image_processor=False, keep_mm_feature_on_device=False, enable_return_hidden_states=False, enable_return_routed_experts=False, scheduler_recv_interval=1, numa_node=None, enable_deterministic_inference=False, rl_on_policy_target=None, enable_attn_tp_input_scattered=False, enable_nsa_prefill_context_parallel=False, nsa_prefill_cp_mode='round-robin-split', enable_fused_qk_norm_rope=False, enable_precise_embedding_interpolation=False, enable_dynamic_batch_tokenizer=False, dynamic_batch_tokenizer_batch_size=32, dynamic_batch_tokenizer_batch_timeout=0.002, debug_tensor_dump_output_folder=None, debug_tensor_dump_layers=None, debug_tensor_dump_input_file=None, debug_tensor_dump_inject=False, disaggregation_mode='null', disaggregation_transfer_backend='mooncake', disaggregation_bootstrap_port=8998, disaggregation_decode_tp=None, disaggregation_decode_dp=None, disaggregation_prefill_pp=1, disaggregation_ib_device=None, disaggregation_decode_enable_offload_kvcache=False, num_reserved_decode_tokens=512, disaggregation_decode_polling_interval=1, encoder_only=False, language_only=False, encoder_transfer_backend='zmq_to_scheduler', encoder_urls=[], custom_weight_loader=[], weight_loader_disable_mmap=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, enable_pdmux=False, pdmux_config_path=None, sm_group_num=8, mm_max_concurrent_calls=32, mm_per_request_timeout=10.0, enable_broadcast_mm_inputs_process=False, enable_prefix_mm_cache=False, mm_enable_dp_encoder=False, mm_process_config={}, limit_mm_data_per_request=None, decrypted_config_file=None, decrypted_draft_config_file=None, forward_hooks=None)
    4. 

  4. [2026-03-27 16:29:04] Ignore import error when loading sglang.srt.multimodal.processors.glmasr: cannot import name 'GlmAsrConfig' from 'transformers' (/usr/local/lib/python3.12/dist-packages/transformers/__init__.py)
    5. 

  5. [2026-03-27 16:29:06] Using default HuggingFace chat template with detected content format: openai
    6. 

  6. [2026-03-27 16:29:12] Mamba selective_state_update backend initialized: triton
    7. 

  7. [2026-03-27 16:29:12] Init torch distributed begin.
    8. 

  8. [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
    9. 

  9. [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
    10. 

  10. [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
    11. 

  11. [2026-03-27 16:29:12] Init torch distributed ends. elapsed=0.24 s, mem usage=0.06 GB
    12. 

  12. [2026-03-27 16:29:12] Ignore import error when loading sglang.srt.models.glm_ocr: No module named 'transformers.models.glm_ocr'
    13. 

  13. [2026-03-27 16:29:12] Ignore import error when loading sglang.srt.models.glm_ocr_nextn: No module named 'transformers.models.glm_ocr'
    14. 

  14. [2026-03-27 16:29:12] Ignore import error when loading sglang.srt.models.glmasr: cannot import name 'GlmAsrConfig' from 'transformers' (/usr/local/lib/python3.12/dist-packages/transformers/__init__.py)
    15. 

  15. [2026-03-27 16:29:13] Load weight begin. avail mem=139.34 GB
    16. 

  16. [2026-03-27 16:29:13] Multimodal attention backend not set. Use fa3.
    17. 

  17. [2026-03-27 16:29:13] Using fa3 as multimodal attention backend.
    18. 

  18. `torch_dtype` is deprecated! Use `dtype` instead!
    19. 

  19. [2026-03-27 16:29:13] using attn output gate!
    20. 

  20. 
Loading safetensors checkpoint shards:   0% Completed | 0/14 [00:00<?, ?it/s]
    21. 

  21. 
Loading safetensors checkpoint shards:   7% Completed | 1/14 [00:00<00:12,  1.06it/s]
    22. 

  22. 
Loading safetensors checkpoint shards:  14% Completed | 2/14 [00:01<00:11,  1.01it/s]
    23. 

  23. 
Loading safetensors checkpoint shards:  21% Completed | 3/14 [00:02<00:11,  1.00s/it]
    24. 

  24. 
Loading safetensors checkpoint shards:  29% Completed | 4/14 [00:04<00:10,  1.01s/it]
    25. 

  25. 
Loading safetensors checkpoint shards:  36% Completed | 5/14 [00:05<00:09,  1.01s/it]
    26. 

  26. 
Loading safetensors checkpoint shards:  43% Completed | 6/14 [00:06<00:08,  1.02s/it]
    27. 

  27. 
Loading safetensors checkpoint shards:  50% Completed | 7/14 [00:07<00:07,  1.02s/it]
    28. 

  28. 
Loading safetensors checkpoint shards:  57% Completed | 8/14 [00:08<00:06,  1.02s/it]
    29. 

  29. 
Loading safetensors checkpoint shards:  64% Completed | 9/14 [00:09<00:04,  1.01it/s]
    30. 

  30. 
Loading safetensors checkpoint shards:  71% Completed | 10/14 [00:09<00:02,  1.34it/s]
    31. 

  31. 
Loading safetensors checkpoint shards:  79% Completed | 11/14 [00:10<00:02,  1.23it/s]
    32. 

  32. 
Loading safetensors checkpoint shards:  86% Completed | 12/14 [00:11<00:01,  1.14it/s]
    33. 

  33. 
Loading safetensors checkpoint shards:  93% Completed | 13/14 [00:12<00:00,  1.09it/s]
    34. 

  34. 
Loading safetensors checkpoint shards: 100% Completed | 14/14 [00:13<00:00,  1.09it/s]
    35. 

  35. 
Loading safetensors checkpoint shards: 100% Completed | 14/14 [00:13<00:00,  1.07it/s]
    36. 

  36. 

  37. [2026-03-27 16:29:26] Load weight end. elapsed=13.37 s, type=Qwen3_5MoeForConditionalGeneration, dtype=torch.bfloat16, avail mem=73.85 GB, mem usage=65.49 GB.
    38. 

  38. [2026-03-27 16:29:26] Using KV cache dtype: torch.bfloat16
    39. 

  39. [2026-03-27 16:29:26] Mamba Cache is allocated. max_mamba_cache_size: 405, conv_state size: 0.56GB, ssm_state size: 23.79GB 
    40. 

  40. [2026-03-27 16:29:26] KV Cache is allocated. #tokens: 1417657, K size: 13.52 GB, V size: 13.52 GB
    41. 

  41. [2026-03-27 16:29:26] Memory pool end. avail mem=22.24 GB
    42. 

  42. [2026-03-27 16:29:27] Using hybrid linear attention backend for hybrid GDN models.
    43. 

  43. [2026-03-27 16:29:27] CuTe DSL GDN decode enabled: False
    44. 

  44. [2026-03-27 16:29:27] Capture cuda graph begin. This can take up to several minutes. avail mem=22.15 GB
    45. 

  45. [2026-03-27 16:29:27] Capture cuda graph bs [1, 2, 4, 8, 12, 16, 24, 32, 40, 48, 56, 64, 72, 80, 88, 96, 104, 112, 120, 128, 135]
    46. 

  46. 
  0%|          | 0/21 [00:00<?, ?it/s]
Capturing batches (bs=135 avail_mem=21.89 GB):   0%|          | 0/21 [00:00<?, ?it/s][2026-03-27 16:29:29] Using default MoE kernel config. Performance might be sub-optimal! Config file not found at /sgl-workspace/sglang/python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=256,N=512,device_name=NVIDIA_H20-3e.json, you can create them with https://github.com/sgl-project/sglang/tree/main/benchmark/kernels/fused_moe_triton
    47. 

  47. [2026-03-27 16:29:29] Using MoE kernel config with down_moe=False. Performance might be sub-optimal! Config file not found at /sgl-workspace/sglang/python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=256,N=512,device_name=NVIDIA_H20-3e_down.json, you can create them with https://github.com/sgl-project/sglang/tree/main/benchmark/kernels/fused_moe_triton
    48. 

  48. 
Capturing batches (bs=135 avail_mem=21.89 GB):   5%|▍         | 1/21 [00:09<03:03,  9.17s/it]
Capturing batches (bs=128 avail_mem=21.74 GB):   5%|▍         | 1/21 [00:09<03:03,  9.17s/it]
Capturing batches (bs=128 avail_mem=21.74 GB):  10%|▉         | 2/21 [00:09<01:19,  4.19s/it]
Capturing batches (bs=120 avail_mem=21.73 GB):  10%|▉         | 2/21 [00:09<01:19,  4.19s/it]
Capturing batches (bs=120 avail_mem=21.73 GB):  14%|█▍        | 3/21 [00:10<00:42,  2.36s/it]
Capturing batches (bs=112 avail_mem=21.73 GB):  14%|█▍        | 3/21 [00:10<00:42,  2.36s/it]
Capturing batches (bs=112 avail_mem=21.73 GB):  19%|█▉        | 4/21 [00:10<00:25,  1.50s/it]
Capturing batches (bs=104 avail_mem=21.73 GB):  19%|█▉        | 4/21 [00:10<00:25,  1.50s/it]
Capturing batches (bs=104 avail_mem=21.73 GB):  24%|██▍       | 5/21 [00:10<00:16,  1.02s/it]
Capturing batches (bs=96 avail_mem=21.72 GB):  24%|██▍       | 5/21 [00:10<00:16,  1.02s/it] 
Capturing batches (bs=96 avail_mem=21.72 GB):  29%|██▊       | 6/21 [00:10<00:11,  1.35it/s]
Capturing batches (bs=88 avail_mem=21.72 GB):  29%|██▊       | 6/21 [00:10<00:11,  1.35it/s]
Capturing batches (bs=88 avail_mem=21.72 GB):  33%|███▎      | 7/21 [00:10<00:07,  1.79it/s]
Capturing batches (bs=80 avail_mem=21.71 GB):  33%|███▎      | 7/21 [00:10<00:07,  1.79it/s]
Capturing batches (bs=80 avail_mem=21.71 GB):  38%|███▊      | 8/21 [00:10<00:05,  2.28it/s]
Capturing batches (bs=72 avail_mem=21.71 GB):  38%|███▊      | 8/21 [00:10<00:05,  2.28it/s]
Capturing batches (bs=72 avail_mem=21.71 GB):  43%|████▎     | 9/21 [00:11<00:04,  2.79it/s]
Capturing batches (bs=64 avail_mem=21.71 GB):  43%|████▎     | 9/21 [00:11<00:04,  2.79it/s]
Capturing batches (bs=64 avail_mem=21.71 GB):  48%|████▊     | 10/21 [00:11<00:03,  3.27it/s]
Capturing batches (bs=56 avail_mem=21.71 GB):  48%|████▊     | 10/21 [00:11<00:03,  3.27it/s]
Capturing batches (bs=56 avail_mem=21.71 GB):  52%|█████▏    | 11/21 [00:11<00:02,  3.75it/s]
Capturing batches (bs=48 avail_mem=21.70 GB):  52%|█████▏    | 11/21 [00:11<00:02,  3.75it/s]
Capturing batches (bs=48 avail_mem=21.70 GB):  57%|█████▋    | 12/21 [00:11<00:02,  4.15it/s]
Capturing batches (bs=40 avail_mem=21.70 GB):  57%|█████▋    | 12/21 [00:11<00:02,  4.15it/s]
Capturing batches (bs=40 avail_mem=21.70 GB):  62%|██████▏   | 13/21 [00:11<00:01,  4.49it/s]
Capturing batches (bs=32 avail_mem=21.69 GB):  62%|██████▏   | 13/21 [00:11<00:01,  4.49it/s]
Capturing batches (bs=32 avail_mem=21.69 GB):  67%|██████▋   | 14/21 [00:13<00:03,  1.80it/s]
Capturing batches (bs=24 avail_mem=21.69 GB):  67%|██████▋   | 14/21 [00:13<00:03,  1.80it/s]
Capturing batches (bs=24 avail_mem=21.69 GB):  71%|███████▏  | 15/21 [00:13<00:03,  1.64it/s]
Capturing batches (bs=16 avail_mem=21.69 GB):  71%|███████▏  | 15/21 [00:13<00:03,  1.64it/s]
Capturing batches (bs=16 avail_mem=21.69 GB):  76%|███████▌  | 16/21 [00:14<00:02,  2.03it/s]
Capturing batches (bs=12 avail_mem=21.68 GB):  76%|███████▌  | 16/21 [00:14<00:02,  2.03it/s]
Capturing batches (bs=12 avail_mem=21.68 GB):  81%|████████  | 17/21 [00:14<00:01,  2.51it/s]
Capturing batches (bs=8 avail_mem=21.68 GB):  81%|████████  | 17/21 [00:14<00:01,  2.51it/s] 
Capturing batches (bs=8 avail_mem=21.68 GB):  86%|████████▌ | 18/21 [00:14<00:01,  2.26it/s]
Capturing batches (bs=4 avail_mem=21.67 GB):  86%|████████▌ | 18/21 [00:14<00:01,  2.26it/s]
Capturing batches (bs=4 avail_mem=21.67 GB):  90%|█████████ | 19/21 [00:15<00:00,  2.47it/s]
Capturing batches (bs=2 avail_mem=21.67 GB):  90%|█████████ | 19/21 [00:15<00:00,  2.47it/s]
Capturing batches (bs=2 avail_mem=21.67 GB):  95%|█████████▌| 20/21 [00:15<00:00,  2.92it/s]
Capturing batches (bs=1 avail_mem=21.66 GB):  95%|█████████▌| 20/21 [00:15<00:00,  2.92it/s]
Capturing batches (bs=1 avail_mem=21.66 GB): 100%|██████████| 21/21 [00:17<00:00,  1.31it/s]
Capturing batches (bs=1 avail_mem=21.66 GB): 100%|██████████| 21/21 [00:17<00:00,  1.22it/s]
    49. 

  49. [2026-03-27 16:29:44] Capture cuda graph end. Time elapsed: 17.75 s. mem usage=0.49 GB. avail mem=21.66 GB.
    50. 

  50. [2026-03-27 16:29:46] max_total_num_tokens=1417657, chunked_prefill_size=8192, max_prefill_tokens=16384, max_running_requests=135, context_len=262144, available_gpu_mem=21.66 GB
    51. 

  51. [2026-03-27 16:29:47] INFO:     Started server process [30]
    52. 

  52. [2026-03-27 16:29:47] INFO:     Waiting for application startup.
    53. 

  53. [2026-03-27 16:29:47] Using default chat sampling params from model generation config: {'repetition_penalty': 1.0, 'temperature': 1.0, 'top_k': 20, 'top_p': 0.95}
    54. 

  54. [2026-03-27 16:29:47] INFO:     Application startup complete.
    55. 

  55. [2026-03-27 16:29:47] INFO:     Uvicorn running on http://0.0.0.0:30000 (Press CTRL+C to quit)
    56. 

  56. [2026-03-27 16:29:48] INFO:     127.0.0.1:42534 - "GET /model_info HTTP/1.1" 200 OK
    57. 

  57. [2026-03-27 16:29:55] Prefill batch, #new-seq: 1, #new-token: 80, #cached-token: 0, full token usage: 0.00, mamba usage: 0.00, #running-req: 0, #queue-req: 0, input throughput (token/s): 0.00, cuda graph: False
    58. 

  58. [2026-03-27 16:29:55] Prefill batch, #new-seq: 1, #new-token: 16, #cached-token: 0, full token usage: 0.00, mamba usage: 0.01, #running-req: 1, #queue-req: 0, input throughput (token/s): 185.02, cuda graph: False
    59. 

  59. [2026-03-27 16:29:56] INFO:     127.0.0.1:42542 - "POST /v1/chat/completions HTTP/1.1" 200 OK
    60. 

  60. [2026-03-27 16:29:56] The server is fired up and ready to roll!
    61. 

  61. [2026-03-27 16:29:56] Decode batch, #running-req: 1, #full token: 56, full token usage: 0.00, mamba num: 2, mamba usage: 0.00, cuda graph: True, gen throughput (token/s): 1.03, #queue-req: 0
    62. 

  62. [2026-03-27 16:29:56] INFO:     172.19.0.1:40908 - "POST /v1/chat/completions HTTP/1.1" 200 OK
    63. 

  63. [2026-03-27 16:30:21] Prefill batch, #new-seq: 1, #new-token: 16, #cached-token: 0, full token usage: 0.00, mamba usage: 0.00, #running-req: 0, #queue-req: 0, input throughput (token/s): 0.63, cuda graph: False
    64. 

  64. [2026-03-27 16:30:21] Decode batch, #running-req: 1, #full token: 47, full token usage: 0.00, mamba num: 2, mamba usage: 0.00, cuda graph: True, gen throughput (token/s): 1.61, #queue-req: 0
    65. 

  65. [2026-03-27 16:30:21] INFO:     172.19.0.1:56604 - "POST /v1/chat/completions HTTP/1.1" 200 OK
    66. 

  66. [2026-03-28 02:30:55] WARNING:  Invalid HTTP request received.
    67. 

  67. [2026-03-28 04:44:49] Prefill batch, #new-seq: 1, #new-token: 16, #cached-token: 0, full token usage: 0.00, mamba usage: 0.00, #running-req: 0, #queue-req: 0, input throughput (token/s): 0.00, cuda graph: False
    68. 

  68. [2026-03-28 04:44:50] Decode batch, #running-req: 1, #full token: 38, full token usage: 0.00, mamba num: 2, mamba usage: 0.00, cuda graph: True, gen throughput (token/s): 0.00, #queue-req: 0
    69. 

  69. [2026-03-28 04:44:50] INFO:     172.19.0.1:60012 - "POST /v1/chat/completions HTTP/1.1" 200 OK
    70.