sglang 0.4.7.post1__py3-none-any.whl → 0.4.8.post1__py3-none-any.whl

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (123) hide show
  1. sglang/bench_one_batch.py +8 -6
  2. sglang/srt/_custom_ops.py +2 -2
  3. sglang/srt/code_completion_parser.py +2 -44
  4. sglang/srt/configs/model_config.py +1 -0
  5. sglang/srt/constants.py +3 -0
  6. sglang/srt/conversation.py +14 -3
  7. sglang/srt/custom_op.py +11 -1
  8. sglang/srt/disaggregation/base/conn.py +2 -0
  9. sglang/srt/disaggregation/decode.py +22 -28
  10. sglang/srt/disaggregation/decode_schedule_batch_mixin.py +4 -3
  11. sglang/srt/disaggregation/mini_lb.py +34 -4
  12. sglang/srt/disaggregation/mooncake/conn.py +301 -64
  13. sglang/srt/disaggregation/mooncake/transfer_engine.py +31 -1
  14. sglang/srt/disaggregation/nixl/conn.py +94 -46
  15. sglang/srt/disaggregation/prefill.py +20 -15
  16. sglang/srt/disaggregation/utils.py +47 -18
  17. sglang/srt/distributed/parallel_state.py +12 -4
  18. sglang/srt/entrypoints/engine.py +27 -31
  19. sglang/srt/entrypoints/http_server.py +149 -79
  20. sglang/srt/entrypoints/http_server_engine.py +0 -3
  21. sglang/srt/entrypoints/openai/__init__.py +0 -0
  22. sglang/srt/{openai_api → entrypoints/openai}/protocol.py +115 -34
  23. sglang/srt/entrypoints/openai/serving_base.py +149 -0
  24. sglang/srt/entrypoints/openai/serving_chat.py +897 -0
  25. sglang/srt/entrypoints/openai/serving_completions.py +425 -0
  26. sglang/srt/entrypoints/openai/serving_embedding.py +170 -0
  27. sglang/srt/entrypoints/openai/serving_rerank.py +102 -0
  28. sglang/srt/entrypoints/openai/serving_score.py +61 -0
  29. sglang/srt/entrypoints/openai/usage_processor.py +81 -0
  30. sglang/srt/entrypoints/openai/utils.py +72 -0
  31. sglang/srt/function_call/base_format_detector.py +7 -4
  32. sglang/srt/function_call/deepseekv3_detector.py +1 -1
  33. sglang/srt/function_call/ebnf_composer.py +64 -10
  34. sglang/srt/function_call/function_call_parser.py +6 -6
  35. sglang/srt/function_call/llama32_detector.py +1 -1
  36. sglang/srt/function_call/mistral_detector.py +1 -1
  37. sglang/srt/function_call/pythonic_detector.py +1 -1
  38. sglang/srt/function_call/qwen25_detector.py +1 -1
  39. sglang/srt/{openai_api/utils.py → jinja_template_utils.py} +6 -5
  40. sglang/srt/layers/activation.py +28 -3
  41. sglang/srt/layers/attention/aiter_backend.py +5 -2
  42. sglang/srt/layers/attention/base_attn_backend.py +1 -1
  43. sglang/srt/layers/attention/cutlass_mla_backend.py +1 -0
  44. sglang/srt/layers/attention/flashattention_backend.py +43 -23
  45. sglang/srt/layers/attention/flashinfer_backend.py +9 -6
  46. sglang/srt/layers/attention/flashinfer_mla_backend.py +7 -4
  47. sglang/srt/layers/attention/flashmla_backend.py +5 -2
  48. sglang/srt/layers/attention/tbo_backend.py +3 -3
  49. sglang/srt/layers/attention/triton_backend.py +19 -11
  50. sglang/srt/layers/communicator.py +5 -5
  51. sglang/srt/layers/dp_attention.py +11 -2
  52. sglang/srt/layers/layernorm.py +44 -2
  53. sglang/srt/layers/linear.py +18 -1
  54. sglang/srt/layers/logits_processor.py +14 -5
  55. sglang/srt/layers/moe/ep_moe/kernels.py +159 -2
  56. sglang/srt/layers/moe/ep_moe/layer.py +286 -13
  57. sglang/srt/layers/moe/ep_moe/token_dispatcher.py +19 -2
  58. sglang/srt/layers/moe/fused_moe_native.py +7 -0
  59. sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_3_1/E=128,N=384,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json +146 -0
  60. sglang/srt/layers/moe/fused_moe_triton/fused_moe.py +13 -2
  61. sglang/srt/layers/moe/fused_moe_triton/layer.py +148 -26
  62. sglang/srt/layers/moe/topk.py +117 -4
  63. sglang/srt/layers/quantization/compressed_tensors/compressed_tensors_moe.py +6 -2
  64. sglang/srt/layers/quantization/fp8.py +25 -17
  65. sglang/srt/layers/quantization/fp8_utils.py +5 -4
  66. sglang/srt/layers/quantization/modelopt_quant.py +62 -8
  67. sglang/srt/layers/quantization/utils.py +5 -2
  68. sglang/srt/layers/rotary_embedding.py +144 -12
  69. sglang/srt/layers/sampler.py +1 -1
  70. sglang/srt/layers/vocab_parallel_embedding.py +14 -1
  71. sglang/srt/lora/lora_manager.py +173 -74
  72. sglang/srt/lora/mem_pool.py +49 -45
  73. sglang/srt/lora/utils.py +1 -1
  74. sglang/srt/managers/cache_controller.py +33 -15
  75. sglang/srt/managers/expert_distribution.py +21 -0
  76. sglang/srt/managers/io_struct.py +19 -14
  77. sglang/srt/managers/multimodal_processors/base_processor.py +44 -9
  78. sglang/srt/managers/multimodal_processors/gemma3n.py +97 -0
  79. sglang/srt/managers/schedule_batch.py +49 -32
  80. sglang/srt/managers/schedule_policy.py +70 -56
  81. sglang/srt/managers/scheduler.py +189 -68
  82. sglang/srt/managers/template_manager.py +226 -0
  83. sglang/srt/managers/tokenizer_manager.py +11 -8
  84. sglang/srt/managers/tp_worker.py +12 -2
  85. sglang/srt/managers/tp_worker_overlap_thread.py +11 -0
  86. sglang/srt/mem_cache/{paged_allocator.py → allocator.py} +125 -34
  87. sglang/srt/mem_cache/base_prefix_cache.py +52 -8
  88. sglang/srt/mem_cache/chunk_cache.py +11 -16
  89. sglang/srt/mem_cache/hiradix_cache.py +34 -23
  90. sglang/srt/mem_cache/memory_pool.py +118 -114
  91. sglang/srt/mem_cache/radix_cache.py +20 -16
  92. sglang/srt/model_executor/cuda_graph_runner.py +77 -46
  93. sglang/srt/model_executor/forward_batch_info.py +18 -5
  94. sglang/srt/model_executor/model_runner.py +27 -8
  95. sglang/srt/model_loader/loader.py +50 -8
  96. sglang/srt/model_loader/weight_utils.py +100 -2
  97. sglang/srt/models/deepseek_nextn.py +35 -30
  98. sglang/srt/models/deepseek_v2.py +255 -30
  99. sglang/srt/models/gemma3n_audio.py +949 -0
  100. sglang/srt/models/gemma3n_causal.py +1009 -0
  101. sglang/srt/models/gemma3n_mm.py +511 -0
  102. sglang/srt/models/glm4.py +312 -0
  103. sglang/srt/models/hunyuan.py +771 -0
  104. sglang/srt/models/mimo_mtp.py +2 -18
  105. sglang/srt/reasoning_parser.py +21 -11
  106. sglang/srt/server_args.py +51 -9
  107. sglang/srt/speculative/eagle_draft_cuda_graph_runner.py +131 -10
  108. sglang/srt/speculative/eagle_draft_extend_cuda_graph_runner.py +125 -12
  109. sglang/srt/speculative/eagle_utils.py +80 -8
  110. sglang/srt/speculative/eagle_worker.py +124 -41
  111. sglang/srt/torch_memory_saver_adapter.py +19 -15
  112. sglang/srt/two_batch_overlap.py +4 -1
  113. sglang/srt/utils.py +248 -11
  114. sglang/test/test_block_fp8_ep.py +1 -0
  115. sglang/test/test_utils.py +1 -0
  116. sglang/version.py +1 -1
  117. {sglang-0.4.7.post1.dist-info → sglang-0.4.8.post1.dist-info}/METADATA +4 -10
  118. {sglang-0.4.7.post1.dist-info → sglang-0.4.8.post1.dist-info}/RECORD +121 -105
  119. sglang/srt/entrypoints/verl_engine.py +0 -179
  120. sglang/srt/openai_api/adapter.py +0 -2148
  121. {sglang-0.4.7.post1.dist-info → sglang-0.4.8.post1.dist-info}/WHEEL +0 -0
  122. {sglang-0.4.7.post1.dist-info → sglang-0.4.8.post1.dist-info}/licenses/LICENSE +0 -0
  123. {sglang-0.4.7.post1.dist-info → sglang-0.4.8.post1.dist-info}/top_level.txt +0 -0