sglang 0.4.7.post1__py3-none-any.whl → 0.4.8__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 (106) 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/constants.py +3 -0
  5. sglang/srt/conversation.py +13 -3
  6. sglang/srt/custom_op.py +5 -1
  7. sglang/srt/disaggregation/decode.py +22 -28
  8. sglang/srt/disaggregation/decode_schedule_batch_mixin.py +4 -3
  9. sglang/srt/disaggregation/mini_lb.py +34 -4
  10. sglang/srt/disaggregation/mooncake/conn.py +12 -16
  11. sglang/srt/disaggregation/prefill.py +17 -13
  12. sglang/srt/disaggregation/utils.py +46 -18
  13. sglang/srt/distributed/parallel_state.py +12 -4
  14. sglang/srt/entrypoints/engine.py +22 -28
  15. sglang/srt/entrypoints/http_server.py +149 -79
  16. sglang/srt/entrypoints/http_server_engine.py +0 -3
  17. sglang/srt/entrypoints/openai/__init__.py +0 -0
  18. sglang/srt/{openai_api → entrypoints/openai}/protocol.py +67 -29
  19. sglang/srt/entrypoints/openai/serving_base.py +149 -0
  20. sglang/srt/entrypoints/openai/serving_chat.py +921 -0
  21. sglang/srt/entrypoints/openai/serving_completions.py +424 -0
  22. sglang/srt/entrypoints/openai/serving_embedding.py +169 -0
  23. sglang/srt/entrypoints/openai/serving_rerank.py +102 -0
  24. sglang/srt/entrypoints/openai/serving_score.py +61 -0
  25. sglang/srt/entrypoints/openai/usage_processor.py +81 -0
  26. sglang/srt/entrypoints/openai/utils.py +72 -0
  27. sglang/srt/function_call/base_format_detector.py +7 -4
  28. sglang/srt/function_call/deepseekv3_detector.py +1 -1
  29. sglang/srt/function_call/ebnf_composer.py +64 -10
  30. sglang/srt/function_call/function_call_parser.py +6 -6
  31. sglang/srt/function_call/llama32_detector.py +1 -1
  32. sglang/srt/function_call/mistral_detector.py +1 -1
  33. sglang/srt/function_call/pythonic_detector.py +1 -1
  34. sglang/srt/function_call/qwen25_detector.py +1 -1
  35. sglang/srt/{openai_api/utils.py → jinja_template_utils.py} +6 -5
  36. sglang/srt/layers/activation.py +21 -3
  37. sglang/srt/layers/attention/aiter_backend.py +5 -2
  38. sglang/srt/layers/attention/base_attn_backend.py +1 -1
  39. sglang/srt/layers/attention/cutlass_mla_backend.py +1 -0
  40. sglang/srt/layers/attention/flashattention_backend.py +19 -9
  41. sglang/srt/layers/attention/flashinfer_backend.py +9 -6
  42. sglang/srt/layers/attention/flashinfer_mla_backend.py +7 -4
  43. sglang/srt/layers/attention/flashmla_backend.py +5 -2
  44. sglang/srt/layers/attention/tbo_backend.py +3 -3
  45. sglang/srt/layers/attention/triton_backend.py +19 -11
  46. sglang/srt/layers/communicator.py +5 -5
  47. sglang/srt/layers/dp_attention.py +11 -2
  48. sglang/srt/layers/layernorm.py +29 -2
  49. sglang/srt/layers/logits_processor.py +2 -2
  50. sglang/srt/layers/moe/ep_moe/kernels.py +159 -2
  51. sglang/srt/layers/moe/ep_moe/layer.py +207 -1
  52. 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
  53. sglang/srt/layers/moe/fused_moe_triton/fused_moe.py +6 -0
  54. sglang/srt/layers/moe/fused_moe_triton/layer.py +75 -12
  55. sglang/srt/layers/moe/topk.py +91 -4
  56. sglang/srt/layers/quantization/compressed_tensors/compressed_tensors_moe.py +6 -2
  57. sglang/srt/layers/quantization/fp8.py +25 -17
  58. sglang/srt/layers/quantization/modelopt_quant.py +62 -8
  59. sglang/srt/layers/quantization/utils.py +5 -2
  60. sglang/srt/layers/rotary_embedding.py +42 -2
  61. sglang/srt/layers/sampler.py +1 -1
  62. sglang/srt/lora/lora_manager.py +173 -74
  63. sglang/srt/lora/mem_pool.py +49 -45
  64. sglang/srt/lora/utils.py +1 -1
  65. sglang/srt/managers/cache_controller.py +33 -15
  66. sglang/srt/managers/io_struct.py +9 -12
  67. sglang/srt/managers/schedule_batch.py +40 -31
  68. sglang/srt/managers/schedule_policy.py +70 -56
  69. sglang/srt/managers/scheduler.py +147 -62
  70. sglang/srt/managers/template_manager.py +226 -0
  71. sglang/srt/managers/tokenizer_manager.py +11 -8
  72. sglang/srt/managers/tp_worker.py +12 -2
  73. sglang/srt/managers/tp_worker_overlap_thread.py +11 -0
  74. sglang/srt/mem_cache/{paged_allocator.py → allocator.py} +125 -34
  75. sglang/srt/mem_cache/base_prefix_cache.py +52 -8
  76. sglang/srt/mem_cache/chunk_cache.py +11 -16
  77. sglang/srt/mem_cache/hiradix_cache.py +34 -23
  78. sglang/srt/mem_cache/memory_pool.py +118 -114
  79. sglang/srt/mem_cache/radix_cache.py +20 -16
  80. sglang/srt/model_executor/cuda_graph_runner.py +76 -45
  81. sglang/srt/model_executor/forward_batch_info.py +18 -5
  82. sglang/srt/model_executor/model_runner.py +22 -6
  83. sglang/srt/model_loader/loader.py +8 -1
  84. sglang/srt/model_loader/weight_utils.py +11 -2
  85. sglang/srt/models/deepseek_nextn.py +29 -27
  86. sglang/srt/models/deepseek_v2.py +108 -26
  87. sglang/srt/models/glm4.py +312 -0
  88. sglang/srt/models/mimo_mtp.py +2 -18
  89. sglang/srt/reasoning_parser.py +21 -11
  90. sglang/srt/server_args.py +36 -8
  91. sglang/srt/speculative/eagle_draft_cuda_graph_runner.py +131 -10
  92. sglang/srt/speculative/eagle_draft_extend_cuda_graph_runner.py +125 -12
  93. sglang/srt/speculative/eagle_utils.py +80 -8
  94. sglang/srt/speculative/eagle_worker.py +124 -41
  95. sglang/srt/torch_memory_saver_adapter.py +19 -15
  96. sglang/srt/utils.py +177 -11
  97. sglang/test/test_block_fp8_ep.py +1 -0
  98. sglang/test/test_utils.py +1 -0
  99. sglang/version.py +1 -1
  100. {sglang-0.4.7.post1.dist-info → sglang-0.4.8.dist-info}/METADATA +4 -10
  101. {sglang-0.4.7.post1.dist-info → sglang-0.4.8.dist-info}/RECORD +104 -93
  102. sglang/srt/entrypoints/verl_engine.py +0 -179
  103. sglang/srt/openai_api/adapter.py +0 -2148
  104. {sglang-0.4.7.post1.dist-info → sglang-0.4.8.dist-info}/WHEEL +0 -0
  105. {sglang-0.4.7.post1.dist-info → sglang-0.4.8.dist-info}/licenses/LICENSE +0 -0
  106. {sglang-0.4.7.post1.dist-info → sglang-0.4.8.dist-info}/top_level.txt +0 -0