sglang 0.5.2rc2__py3-none-any.whl → 0.5.3rc0__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 (238) hide show
  1. sglang/bench_one_batch_server.py +10 -1
  2. sglang/bench_serving.py +257 -29
  3. sglang/srt/configs/__init__.py +4 -0
  4. sglang/srt/configs/device_config.py +3 -1
  5. sglang/srt/configs/dots_vlm.py +139 -0
  6. sglang/srt/configs/load_config.py +1 -0
  7. sglang/srt/configs/model_config.py +50 -6
  8. sglang/srt/configs/qwen3_next.py +326 -0
  9. sglang/srt/connector/__init__.py +8 -1
  10. sglang/srt/connector/remote_instance.py +82 -0
  11. sglang/srt/constrained/base_grammar_backend.py +48 -12
  12. sglang/srt/constrained/llguidance_backend.py +0 -1
  13. sglang/srt/constrained/outlines_backend.py +0 -1
  14. sglang/srt/constrained/xgrammar_backend.py +28 -9
  15. sglang/srt/custom_op.py +11 -1
  16. sglang/srt/debug_utils/dump_comparator.py +81 -44
  17. sglang/srt/debug_utils/dump_loader.py +97 -0
  18. sglang/srt/debug_utils/dumper.py +11 -3
  19. sglang/srt/debug_utils/text_comparator.py +73 -11
  20. sglang/srt/disaggregation/base/conn.py +1 -1
  21. sglang/srt/disaggregation/common/conn.py +15 -12
  22. sglang/srt/disaggregation/decode.py +21 -10
  23. sglang/srt/disaggregation/decode_schedule_batch_mixin.py +4 -1
  24. sglang/srt/disaggregation/fake/conn.py +1 -1
  25. sglang/srt/disaggregation/mini_lb.py +6 -445
  26. sglang/srt/disaggregation/mooncake/conn.py +18 -10
  27. sglang/srt/disaggregation/nixl/conn.py +180 -16
  28. sglang/srt/disaggregation/prefill.py +5 -3
  29. sglang/srt/disaggregation/utils.py +5 -50
  30. sglang/srt/distributed/parallel_state.py +24 -3
  31. sglang/srt/entrypoints/engine.py +38 -17
  32. sglang/srt/entrypoints/grpc_request_manager.py +580 -0
  33. sglang/srt/entrypoints/grpc_server.py +680 -0
  34. sglang/srt/entrypoints/http_server.py +85 -54
  35. sglang/srt/entrypoints/openai/protocol.py +4 -1
  36. sglang/srt/entrypoints/openai/serving_base.py +46 -3
  37. sglang/srt/entrypoints/openai/serving_chat.py +36 -16
  38. sglang/srt/entrypoints/openai/serving_completions.py +12 -3
  39. sglang/srt/entrypoints/openai/serving_embedding.py +8 -3
  40. sglang/srt/entrypoints/openai/serving_rerank.py +3 -1
  41. sglang/srt/entrypoints/openai/serving_responses.py +6 -3
  42. sglang/srt/entrypoints/openai/serving_score.py +1 -0
  43. sglang/srt/eplb/eplb_manager.py +2 -2
  44. sglang/srt/eplb/expert_distribution.py +26 -13
  45. sglang/srt/eplb/expert_location.py +8 -3
  46. sglang/srt/eplb/expert_location_updater.py +1 -1
  47. sglang/srt/function_call/base_format_detector.py +3 -6
  48. sglang/srt/function_call/ebnf_composer.py +11 -9
  49. sglang/srt/function_call/function_call_parser.py +6 -0
  50. sglang/srt/function_call/glm4_moe_detector.py +1 -1
  51. sglang/srt/function_call/qwen3_coder_detector.py +1 -1
  52. sglang/srt/grpc/__init__.py +1 -0
  53. sglang/srt/grpc/sglang_scheduler_pb2.py +106 -0
  54. sglang/srt/grpc/sglang_scheduler_pb2.pyi +427 -0
  55. sglang/srt/grpc/sglang_scheduler_pb2_grpc.py +236 -0
  56. sglang/srt/hf_transformers_utils.py +4 -0
  57. sglang/srt/layers/activation.py +142 -9
  58. sglang/srt/layers/attention/ascend_backend.py +11 -4
  59. sglang/srt/layers/attention/fla/chunk.py +242 -0
  60. sglang/srt/layers/attention/fla/chunk_delta_h.py +314 -0
  61. sglang/srt/layers/attention/fla/chunk_o.py +178 -0
  62. sglang/srt/layers/attention/fla/chunk_scaled_dot_kkt.py +151 -0
  63. sglang/srt/layers/attention/fla/cumsum.py +300 -0
  64. sglang/srt/layers/attention/fla/fused_recurrent.py +640 -0
  65. sglang/srt/layers/attention/fla/fused_sigmoid_gating_recurrent.py +232 -0
  66. sglang/srt/layers/attention/fla/index.py +37 -0
  67. sglang/srt/layers/attention/fla/l2norm.py +150 -0
  68. sglang/srt/layers/attention/fla/layernorm_gated.py +326 -0
  69. sglang/srt/layers/attention/fla/op.py +66 -0
  70. sglang/srt/layers/attention/fla/solve_tril.py +465 -0
  71. sglang/srt/layers/attention/fla/utils.py +331 -0
  72. sglang/srt/layers/attention/fla/wy_fast.py +158 -0
  73. sglang/srt/layers/attention/flashinfer_backend.py +6 -4
  74. sglang/srt/layers/attention/flashinfer_mla_backend.py +16 -12
  75. sglang/srt/layers/attention/hybrid_attn_backend.py +57 -50
  76. sglang/srt/layers/attention/hybrid_linear_attn_backend.py +602 -0
  77. sglang/srt/layers/attention/intel_amx_backend.py +3 -0
  78. sglang/srt/layers/attention/mamba/causal_conv1d.py +128 -0
  79. sglang/srt/layers/attention/mamba/causal_conv1d_triton.py +1052 -0
  80. sglang/srt/layers/attention/mamba/mamba.py +64 -0
  81. sglang/srt/layers/attention/torch_native_backend.py +12 -6
  82. sglang/srt/layers/attention/triton_backend.py +18 -1
  83. sglang/srt/layers/attention/trtllm_mla_backend.py +124 -31
  84. sglang/srt/layers/attention/wave_ops/decode_attention.py +2 -4
  85. sglang/srt/layers/attention/wave_ops/extend_attention.py +1 -3
  86. sglang/srt/layers/dp_attention.py +30 -1
  87. sglang/srt/layers/layernorm.py +32 -15
  88. sglang/srt/layers/linear.py +34 -3
  89. sglang/srt/layers/logits_processor.py +29 -10
  90. sglang/srt/layers/moe/__init__.py +2 -1
  91. sglang/srt/layers/moe/cutlass_w4a8_moe.py +3 -3
  92. sglang/srt/layers/moe/ep_moe/kernels.py +1 -1
  93. sglang/srt/layers/moe/ep_moe/layer.py +182 -62
  94. sglang/srt/layers/moe/flashinfer_cutedsl_moe.py +156 -0
  95. sglang/srt/layers/moe/fused_moe_native.py +5 -3
  96. sglang/srt/layers/moe/fused_moe_triton/configs/{triton_3_4_0/E=128,N=768,device_name=NVIDIA_B200,dtype=fp8_w8a8,block_shape=[128, 128].json → triton_3_3_1/E=128,N=768,device_name=NVIDIA_H20,dtype=fp8_w8a8,block_shape=[128, 128].json } +35 -35
  97. sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_4_0/E=128,N=352,device_name=NVIDIA_RTX_5880_Ada_Generation,dtype=fp8_w8a8.json +146 -0
  98. sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_4_0/E=256,N=512,device_name=NVIDIA_H20.json +146 -0
  99. sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_4_0/E=512,N=128,device_name=NVIDIA_H100_80GB_HBM3.json +146 -0
  100. sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_4_0/E=512,N=128,device_name=NVIDIA_H20-3e.json +146 -0
  101. sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_4_0/E=512,N=128,device_name=NVIDIA_H200.json +146 -0
  102. sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_4_0/E=512,N=256,device_name=NVIDIA_H20-3e.json +146 -0
  103. sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_4_0/E=512,N=256,device_name=NVIDIA_H200.json +146 -0
  104. sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_4_0/E=512,N=64,device_name=NVIDIA_H100_80GB_HBM3.json +146 -0
  105. sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_4_0/E=512,N=64,device_name=NVIDIA_H200.json +146 -0
  106. sglang/srt/layers/moe/fused_moe_triton/fused_moe.py +5 -2
  107. sglang/srt/layers/moe/fused_moe_triton/fused_moe_triton_config.py +1 -1
  108. sglang/srt/layers/moe/fused_moe_triton/fused_moe_triton_kernels.py +23 -20
  109. sglang/srt/layers/moe/fused_moe_triton/layer.py +61 -59
  110. sglang/srt/layers/moe/moe_runner/__init__.py +2 -1
  111. sglang/srt/layers/moe/moe_runner/base.py +274 -1
  112. sglang/srt/layers/moe/moe_runner/runner.py +80 -0
  113. sglang/srt/layers/moe/moe_runner/triton.py +448 -0
  114. sglang/srt/layers/moe/token_dispatcher/__init__.py +16 -4
  115. sglang/srt/layers/moe/token_dispatcher/{base_dispatcher.py → base.py} +67 -17
  116. sglang/srt/layers/moe/token_dispatcher/deepep.py +43 -39
  117. sglang/srt/layers/moe/token_dispatcher/standard.py +44 -2
  118. sglang/srt/layers/moe/topk.py +30 -9
  119. sglang/srt/layers/moe/utils.py +12 -6
  120. sglang/srt/layers/quantization/awq.py +19 -7
  121. sglang/srt/layers/quantization/base_config.py +11 -6
  122. sglang/srt/layers/quantization/blockwise_int8.py +38 -27
  123. sglang/srt/layers/quantization/compressed_tensors/compressed_tensors_moe.py +50 -30
  124. sglang/srt/layers/quantization/compressed_tensors/schemes/compressed_tensors_w8a8_fp8.py +13 -1
  125. sglang/srt/layers/quantization/deep_gemm_wrapper/entrypoint.py +27 -0
  126. sglang/srt/layers/quantization/fp8.py +76 -47
  127. sglang/srt/layers/quantization/fp8_utils.py +50 -31
  128. sglang/srt/layers/quantization/gptq.py +25 -17
  129. sglang/srt/layers/quantization/modelopt_quant.py +147 -47
  130. sglang/srt/layers/quantization/moe_wna16.py +21 -18
  131. sglang/srt/layers/quantization/mxfp4.py +64 -40
  132. sglang/srt/layers/quantization/quark/quark_moe.py +32 -27
  133. sglang/srt/layers/quantization/unquant.py +135 -47
  134. sglang/srt/layers/quantization/w4afp8.py +30 -17
  135. sglang/srt/layers/quantization/w8a8_fp8.py +35 -20
  136. sglang/srt/layers/quantization/w8a8_int8.py +76 -38
  137. sglang/srt/layers/sampler.py +162 -18
  138. sglang/srt/lora/backend/base_backend.py +50 -8
  139. sglang/srt/lora/backend/triton_backend.py +90 -2
  140. sglang/srt/lora/layers.py +32 -0
  141. sglang/srt/lora/lora.py +4 -1
  142. sglang/srt/lora/lora_manager.py +35 -112
  143. sglang/srt/lora/mem_pool.py +24 -10
  144. sglang/srt/lora/utils.py +18 -9
  145. sglang/srt/managers/async_dynamic_batch_tokenizer.py +170 -0
  146. sglang/srt/managers/cache_controller.py +158 -160
  147. sglang/srt/managers/data_parallel_controller.py +105 -35
  148. sglang/srt/managers/detokenizer_manager.py +8 -4
  149. sglang/srt/managers/disagg_service.py +46 -0
  150. sglang/srt/managers/io_struct.py +199 -12
  151. sglang/srt/managers/mm_utils.py +1 -0
  152. sglang/srt/managers/multi_tokenizer_mixin.py +350 -400
  153. sglang/srt/managers/schedule_batch.py +77 -56
  154. sglang/srt/managers/schedule_policy.py +1 -1
  155. sglang/srt/managers/scheduler.py +187 -39
  156. sglang/srt/managers/scheduler_metrics_mixin.py +4 -3
  157. sglang/srt/managers/scheduler_output_processor_mixin.py +55 -11
  158. sglang/srt/managers/scheduler_profiler_mixin.py +1 -1
  159. sglang/srt/managers/tokenizer_communicator_mixin.py +569 -0
  160. sglang/srt/managers/tokenizer_manager.py +259 -519
  161. sglang/srt/managers/tp_worker.py +53 -4
  162. sglang/srt/managers/tp_worker_overlap_thread.py +42 -19
  163. sglang/srt/mem_cache/hicache_storage.py +3 -23
  164. sglang/srt/mem_cache/hiradix_cache.py +103 -43
  165. sglang/srt/mem_cache/memory_pool.py +347 -48
  166. sglang/srt/mem_cache/memory_pool_host.py +105 -46
  167. sglang/srt/mem_cache/radix_cache.py +0 -2
  168. sglang/srt/mem_cache/storage/hf3fs/hf3fs_client.py +164 -0
  169. sglang/srt/mem_cache/storage/hf3fs/{client_hf3fs.py → hf3fs_usrbio_client.py} +5 -1
  170. sglang/srt/mem_cache/storage/hf3fs/storage_hf3fs.py +86 -4
  171. sglang/srt/mem_cache/storage/lmcache/lmc_radix_cache.py +280 -0
  172. sglang/srt/mem_cache/storage/lmcache/unit_test.py +121 -0
  173. sglang/srt/mem_cache/storage/mooncake_store/mooncake_store.py +49 -7
  174. sglang/srt/mem_cache/swa_radix_cache.py +0 -2
  175. sglang/srt/metrics/collector.py +493 -76
  176. sglang/srt/metrics/startup_func_log_and_timer.py +150 -0
  177. sglang/srt/model_executor/cpu_graph_runner.py +640 -0
  178. sglang/srt/model_executor/cuda_graph_runner.py +13 -5
  179. sglang/srt/model_executor/forward_batch_info.py +59 -2
  180. sglang/srt/model_executor/model_runner.py +356 -29
  181. sglang/srt/model_loader/__init__.py +9 -3
  182. sglang/srt/model_loader/loader.py +128 -4
  183. sglang/srt/model_loader/weight_utils.py +2 -1
  184. sglang/srt/models/apertus.py +686 -0
  185. sglang/srt/models/bailing_moe.py +798 -218
  186. sglang/srt/models/bailing_moe_nextn.py +168 -0
  187. sglang/srt/models/deepseek_v2.py +109 -15
  188. sglang/srt/models/dots_vlm.py +174 -0
  189. sglang/srt/models/dots_vlm_vit.py +337 -0
  190. sglang/srt/models/ernie4.py +1 -1
  191. sglang/srt/models/gemma3n_mm.py +1 -1
  192. sglang/srt/models/glm4_moe.py +1 -1
  193. sglang/srt/models/glm4v.py +4 -2
  194. sglang/srt/models/glm4v_moe.py +3 -0
  195. sglang/srt/models/gpt_oss.py +1 -1
  196. sglang/srt/models/llama4.py +9 -0
  197. sglang/srt/models/llama_eagle3.py +13 -0
  198. sglang/srt/models/longcat_flash.py +2 -2
  199. sglang/srt/models/mllama4.py +25 -0
  200. sglang/srt/models/opt.py +637 -0
  201. sglang/srt/models/qwen2.py +7 -0
  202. sglang/srt/models/qwen2_5_vl.py +27 -3
  203. sglang/srt/models/qwen2_moe.py +56 -12
  204. sglang/srt/models/qwen3_moe.py +1 -1
  205. sglang/srt/models/qwen3_next.py +1042 -0
  206. sglang/srt/models/qwen3_next_mtp.py +112 -0
  207. sglang/srt/models/step3_vl.py +1 -1
  208. sglang/srt/multimodal/processors/dots_vlm.py +99 -0
  209. sglang/srt/multimodal/processors/glm4v.py +9 -9
  210. sglang/srt/multimodal/processors/internvl.py +141 -129
  211. sglang/srt/multimodal/processors/qwen_vl.py +15 -5
  212. sglang/srt/offloader.py +27 -3
  213. sglang/srt/remote_instance_weight_loader_utils.py +69 -0
  214. sglang/srt/sampling/sampling_batch_info.py +18 -15
  215. sglang/srt/server_args.py +276 -35
  216. sglang/srt/speculative/eagle_draft_cuda_graph_runner.py +5 -0
  217. sglang/srt/speculative/eagle_draft_extend_cuda_graph_runner.py +10 -1
  218. sglang/srt/speculative/eagle_utils.py +0 -2
  219. sglang/srt/speculative/eagle_worker.py +43 -4
  220. sglang/srt/speculative/spec_info.py +5 -0
  221. sglang/srt/speculative/standalone_worker.py +109 -0
  222. sglang/srt/tracing/trace.py +552 -0
  223. sglang/srt/utils.py +34 -3
  224. sglang/srt/weight_sync/utils.py +1 -1
  225. sglang/test/attention/test_trtllm_mla_backend.py +169 -5
  226. sglang/test/runners.py +4 -0
  227. sglang/test/test_cutlass_moe.py +24 -6
  228. sglang/test/test_disaggregation_utils.py +66 -0
  229. sglang/test/test_fp4_moe.py +370 -1
  230. sglang/test/test_utils.py +28 -1
  231. sglang/utils.py +11 -0
  232. sglang/version.py +1 -1
  233. {sglang-0.5.2rc2.dist-info → sglang-0.5.3rc0.dist-info}/METADATA +59 -123
  234. {sglang-0.5.2rc2.dist-info → sglang-0.5.3rc0.dist-info}/RECORD +237 -178
  235. sglang/srt/disaggregation/launch_lb.py +0 -118
  236. {sglang-0.5.2rc2.dist-info → sglang-0.5.3rc0.dist-info}/WHEEL +0 -0
  237. {sglang-0.5.2rc2.dist-info → sglang-0.5.3rc0.dist-info}/licenses/LICENSE +0 -0
  238. {sglang-0.5.2rc2.dist-info → sglang-0.5.3rc0.dist-info}/top_level.txt +0 -0
sglang/srt/layers/moe/fused_moe_triton/configs/{triton_3_4_0/E=128,N=768,device_name=NVIDIA_B200,dtype=fp8_w8a8,block_shape=[128, 128].json → triton_3_3_1/E=128,N=768,device_name=NVIDIA_H20,dtype=fp8_w8a8,block_shape=[128, 128].json } RENAMED
@@ -1,27 +1,27 @@
1
1
  {
2
2
  "1": {
3
3
  "BLOCK_SIZE_M": 16,
4
- "BLOCK_SIZE_N": 128,
4
+ "BLOCK_SIZE_N": 64,
5
5
  "BLOCK_SIZE_K": 128,
6
6
  "GROUP_SIZE_M": 1,
7
7
  "num_warps": 4,
8
- "num_stages": 4
8
+ "num_stages": 5
9
9
  },
10
10
  "2": {
11
11
  "BLOCK_SIZE_M": 16,
12
- "BLOCK_SIZE_N": 256,
12
+ "BLOCK_SIZE_N": 64,
13
13
  "BLOCK_SIZE_K": 128,
14
14
  "GROUP_SIZE_M": 1,
15
15
  "num_warps": 4,
16
- "num_stages": 5
16
+ "num_stages": 3
17
17
  },
18
18
  "4": {
19
19
  "BLOCK_SIZE_M": 16,
20
20
  "BLOCK_SIZE_N": 128,
21
- "BLOCK_SIZE_K": 256,
22
- "GROUP_SIZE_M": 1,
21
+ "BLOCK_SIZE_K": 128,
22
+ "GROUP_SIZE_M": 32,
23
23
  "num_warps": 4,
24
- "num_stages": 4
24
+ "num_stages": 3
25
25
  },
26
26
  "8": {
27
27
  "BLOCK_SIZE_M": 16,
@@ -29,31 +29,31 @@
29
29
  "BLOCK_SIZE_K": 128,
30
30
  "GROUP_SIZE_M": 1,
31
31
  "num_warps": 4,
32
- "num_stages": 4
32
+ "num_stages": 3
33
33
  },
34
34
  "16": {
35
35
  "BLOCK_SIZE_M": 16,
36
36
  "BLOCK_SIZE_N": 128,
37
- "BLOCK_SIZE_K": 256,
37
+ "BLOCK_SIZE_K": 128,
38
38
  "GROUP_SIZE_M": 1,
39
39
  "num_warps": 4,
40
- "num_stages": 4
40
+ "num_stages": 3
41
41
  },
42
42
  "24": {
43
43
  "BLOCK_SIZE_M": 16,
44
44
  "BLOCK_SIZE_N": 128,
45
- "BLOCK_SIZE_K": 256,
45
+ "BLOCK_SIZE_K": 128,
46
46
  "GROUP_SIZE_M": 1,
47
47
  "num_warps": 4,
48
- "num_stages": 4
48
+ "num_stages": 3
49
49
  },
50
50
  "32": {
51
51
  "BLOCK_SIZE_M": 16,
52
52
  "BLOCK_SIZE_N": 128,
53
- "BLOCK_SIZE_K": 256,
53
+ "BLOCK_SIZE_K": 128,
54
54
  "GROUP_SIZE_M": 1,
55
55
  "num_warps": 4,
56
- "num_stages": 4
56
+ "num_stages": 3
57
57
  },
58
58
  "48": {
59
59
  "BLOCK_SIZE_M": 16,
@@ -61,55 +61,55 @@
61
61
  "BLOCK_SIZE_K": 128,
62
62
  "GROUP_SIZE_M": 1,
63
63
  "num_warps": 4,
64
- "num_stages": 4
64
+ "num_stages": 3
65
65
  },
66
66
  "64": {
67
67
  "BLOCK_SIZE_M": 16,
68
68
  "BLOCK_SIZE_N": 128,
69
- "BLOCK_SIZE_K": 256,
69
+ "BLOCK_SIZE_K": 128,
70
70
  "GROUP_SIZE_M": 1,
71
71
  "num_warps": 4,
72
- "num_stages": 4
72
+ "num_stages": 3
73
73
  },
74
74
  "96": {
75
75
  "BLOCK_SIZE_M": 16,
76
76
  "BLOCK_SIZE_N": 128,
77
- "BLOCK_SIZE_K": 256,
77
+ "BLOCK_SIZE_K": 128,
78
78
  "GROUP_SIZE_M": 1,
79
79
  "num_warps": 4,
80
- "num_stages": 4
80
+ "num_stages": 3
81
81
  },
82
82
  "128": {
83
83
  "BLOCK_SIZE_M": 16,
84
84
  "BLOCK_SIZE_N": 128,
85
- "BLOCK_SIZE_K": 256,
85
+ "BLOCK_SIZE_K": 128,
86
86
  "GROUP_SIZE_M": 1,
87
87
  "num_warps": 4,
88
- "num_stages": 4
88
+ "num_stages": 3
89
89
  },
90
90
  "256": {
91
- "BLOCK_SIZE_M": 16,
91
+ "BLOCK_SIZE_M": 64,
92
92
  "BLOCK_SIZE_N": 128,
93
93
  "BLOCK_SIZE_K": 128,
94
- "GROUP_SIZE_M": 1,
94
+ "GROUP_SIZE_M": 32,
95
95
  "num_warps": 4,
96
96
  "num_stages": 3
97
97
  },
98
98
  "512": {
99
99
  "BLOCK_SIZE_M": 64,
100
100
  "BLOCK_SIZE_N": 128,
101
- "BLOCK_SIZE_K": 256,
101
+ "BLOCK_SIZE_K": 128,
102
102
  "GROUP_SIZE_M": 32,
103
103
  "num_warps": 4,
104
- "num_stages": 4
104
+ "num_stages": 3
105
105
  },
106
106
  "1024": {
107
107
  "BLOCK_SIZE_M": 64,
108
108
  "BLOCK_SIZE_N": 128,
109
109
  "BLOCK_SIZE_K": 128,
110
- "GROUP_SIZE_M": 32,
110
+ "GROUP_SIZE_M": 1,
111
111
  "num_warps": 4,
112
- "num_stages": 4
112
+ "num_stages": 3
113
113
  },
114
114
  "1536": {
115
115
  "BLOCK_SIZE_M": 64,
@@ -122,25 +122,25 @@
122
122
  "2048": {
123
123
  "BLOCK_SIZE_M": 64,
124
124
  "BLOCK_SIZE_N": 128,
125
- "BLOCK_SIZE_K": 256,
126
- "GROUP_SIZE_M": 32,
125
+ "BLOCK_SIZE_K": 128,
126
+ "GROUP_SIZE_M": 16,
127
127
  "num_warps": 4,
128
- "num_stages": 4
128
+ "num_stages": 3
129
129
  },
130
130
  "3072": {
131
131
  "BLOCK_SIZE_M": 64,
132
132
  "BLOCK_SIZE_N": 128,
133
133
  "BLOCK_SIZE_K": 128,
134
- "GROUP_SIZE_M": 32,
134
+ "GROUP_SIZE_M": 1,
135
135
  "num_warps": 4,
136
- "num_stages": 4
136
+ "num_stages": 3
137
137
  },
138
138
  "4096": {
139
139
  "BLOCK_SIZE_M": 64,
140
140
  "BLOCK_SIZE_N": 128,
141
- "BLOCK_SIZE_K": 256,
142
- "GROUP_SIZE_M": 16,
141
+ "BLOCK_SIZE_K": 128,
142
+ "GROUP_SIZE_M": 1,
143
143
  "num_warps": 4,
144
- "num_stages": 4
144
+ "num_stages": 3
145
145
  }
146
146
  }
sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_4_0/E=128,N=352,device_name=NVIDIA_RTX_5880_Ada_Generation,dtype=fp8_w8a8.json ADDED
@@ -0,0 +1,146 @@
1
+ {
2
+ "1": {
3
+ "BLOCK_SIZE_M": 16,
4
+ "BLOCK_SIZE_N": 64,
5
+ "BLOCK_SIZE_K": 128,
6
+ "GROUP_SIZE_M": 1,
7
+ "num_warps": 4,
8
+ "num_stages": 2
9
+ },
10
+ "2": {
11
+ "BLOCK_SIZE_M": 16,
12
+ "BLOCK_SIZE_N": 64,
13
+ "BLOCK_SIZE_K": 128,
14
+ "GROUP_SIZE_M": 1,
15
+ "num_warps": 4,
16
+ "num_stages": 2
17
+ },
18
+ "4": {
19
+ "BLOCK_SIZE_M": 16,
20
+ "BLOCK_SIZE_N": 128,
21
+ "BLOCK_SIZE_K": 128,
22
+ "GROUP_SIZE_M": 16,
23
+ "num_warps": 8,
24
+ "num_stages": 5
25
+ },
26
+ "8": {
27
+ "BLOCK_SIZE_M": 32,
28
+ "BLOCK_SIZE_N": 128,
29
+ "BLOCK_SIZE_K": 256,
30
+ "GROUP_SIZE_M": 32,
31
+ "num_warps": 8,
32
+ "num_stages": 2
33
+ },
34
+ "16": {
35
+ "BLOCK_SIZE_M": 16,
36
+ "BLOCK_SIZE_N": 64,
37
+ "BLOCK_SIZE_K": 256,
38
+ "GROUP_SIZE_M": 64,
39
+ "num_warps": 8,
40
+ "num_stages": 2
41
+ },
42
+ "24": {
43
+ "BLOCK_SIZE_M": 16,
44
+ "BLOCK_SIZE_N": 64,
45
+ "BLOCK_SIZE_K": 256,
46
+ "GROUP_SIZE_M": 32,
47
+ "num_warps": 4,
48
+ "num_stages": 2
49
+ },
50
+ "32": {
51
+ "BLOCK_SIZE_M": 16,
52
+ "BLOCK_SIZE_N": 32,
53
+ "BLOCK_SIZE_K": 256,
54
+ "GROUP_SIZE_M": 1,
55
+ "num_warps": 8,
56
+ "num_stages": 3
57
+ },
58
+ "48": {
59
+ "BLOCK_SIZE_M": 16,
60
+ "BLOCK_SIZE_N": 32,
61
+ "BLOCK_SIZE_K": 256,
62
+ "GROUP_SIZE_M": 32,
63
+ "num_warps": 4,
64
+ "num_stages": 2
65
+ },
66
+ "64": {
67
+ "BLOCK_SIZE_M": 16,
68
+ "BLOCK_SIZE_N": 32,
69
+ "BLOCK_SIZE_K": 128,
70
+ "GROUP_SIZE_M": 16,
71
+ "num_warps": 4,
72
+ "num_stages": 5
73
+ },
74
+ "96": {
75
+ "BLOCK_SIZE_M": 16,
76
+ "BLOCK_SIZE_N": 64,
77
+ "BLOCK_SIZE_K": 256,
78
+ "GROUP_SIZE_M": 1,
79
+ "num_warps": 8,
80
+ "num_stages": 2
81
+ },
82
+ "128": {
83
+ "BLOCK_SIZE_M": 16,
84
+ "BLOCK_SIZE_N": 64,
85
+ "BLOCK_SIZE_K": 256,
86
+ "GROUP_SIZE_M": 16,
87
+ "num_warps": 8,
88
+ "num_stages": 2
89
+ },
90
+ "256": {
91
+ "BLOCK_SIZE_M": 32,
92
+ "BLOCK_SIZE_N": 64,
93
+ "BLOCK_SIZE_K": 256,
94
+ "GROUP_SIZE_M": 1,
95
+ "num_warps": 8,
96
+ "num_stages": 2
97
+ },
98
+ "512": {
99
+ "BLOCK_SIZE_M": 32,
100
+ "BLOCK_SIZE_N": 64,
101
+ "BLOCK_SIZE_K": 256,
102
+ "GROUP_SIZE_M": 64,
103
+ "num_warps": 4,
104
+ "num_stages": 2
105
+ },
106
+ "1024": {
107
+ "BLOCK_SIZE_M": 64,
108
+ "BLOCK_SIZE_N": 64,
109
+ "BLOCK_SIZE_K": 128,
110
+ "GROUP_SIZE_M": 32,
111
+ "num_warps": 4,
112
+ "num_stages": 3
113
+ },
114
+ "1536": {
115
+ "BLOCK_SIZE_M": 128,
116
+ "BLOCK_SIZE_N": 128,
117
+ "BLOCK_SIZE_K": 128,
118
+ "GROUP_SIZE_M": 16,
119
+ "num_warps": 8,
120
+ "num_stages": 4
121
+ },
122
+ "2048": {
123
+ "BLOCK_SIZE_M": 128,
124
+ "BLOCK_SIZE_N": 256,
125
+ "BLOCK_SIZE_K": 128,
126
+ "GROUP_SIZE_M": 1,
127
+ "num_warps": 8,
128
+ "num_stages": 3
129
+ },
130
+ "3072": {
131
+ "BLOCK_SIZE_M": 128,
132
+ "BLOCK_SIZE_N": 256,
133
+ "BLOCK_SIZE_K": 128,
134
+ "GROUP_SIZE_M": 1,
135
+ "num_warps": 8,
136
+ "num_stages": 3
137
+ },
138
+ "4096": {
139
+ "BLOCK_SIZE_M": 128,
140
+ "BLOCK_SIZE_N": 256,
141
+ "BLOCK_SIZE_K": 128,
142
+ "GROUP_SIZE_M": 1,
143
+ "num_warps": 8,
144
+ "num_stages": 3
145
+ }
146
+ }
sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_4_0/E=256,N=512,device_name=NVIDIA_H20.json ADDED
@@ -0,0 +1,146 @@
1
+ {
2
+ "1": {
3
+ "BLOCK_SIZE_M": 16,
4
+ "BLOCK_SIZE_N": 64,
5
+ "BLOCK_SIZE_K": 64,
6
+ "GROUP_SIZE_M": 1,
7
+ "num_warps": 4,
8
+ "num_stages": 4
9
+ },
10
+ "2": {
11
+ "BLOCK_SIZE_M": 16,
12
+ "BLOCK_SIZE_N": 64,
13
+ "BLOCK_SIZE_K": 64,
14
+ "GROUP_SIZE_M": 1,
15
+ "num_warps": 4,
16
+ "num_stages": 3
17
+ },
18
+ "4": {
19
+ "BLOCK_SIZE_M": 16,
20
+ "BLOCK_SIZE_N": 64,
21
+ "BLOCK_SIZE_K": 128,
22
+ "GROUP_SIZE_M": 16,
23
+ "num_warps": 4,
24
+ "num_stages": 2
25
+ },
26
+ "8": {
27
+ "BLOCK_SIZE_M": 16,
28
+ "BLOCK_SIZE_N": 64,
29
+ "BLOCK_SIZE_K": 64,
30
+ "GROUP_SIZE_M": 64,
31
+ "num_warps": 4,
32
+ "num_stages": 4
33
+ },
34
+ "16": {
35
+ "BLOCK_SIZE_M": 16,
36
+ "BLOCK_SIZE_N": 128,
37
+ "BLOCK_SIZE_K": 64,
38
+ "GROUP_SIZE_M": 1,
39
+ "num_warps": 8,
40
+ "num_stages": 5
41
+ },
42
+ "24": {
43
+ "BLOCK_SIZE_M": 16,
44
+ "BLOCK_SIZE_N": 128,
45
+ "BLOCK_SIZE_K": 64,
46
+ "GROUP_SIZE_M": 1,
47
+ "num_warps": 8,
48
+ "num_stages": 5
49
+ },
50
+ "32": {
51
+ "BLOCK_SIZE_M": 16,
52
+ "BLOCK_SIZE_N": 128,
53
+ "BLOCK_SIZE_K": 64,
54
+ "GROUP_SIZE_M": 1,
55
+ "num_warps": 8,
56
+ "num_stages": 5
57
+ },
58
+ "48": {
59
+ "BLOCK_SIZE_M": 16,
60
+ "BLOCK_SIZE_N": 128,
61
+ "BLOCK_SIZE_K": 64,
62
+ "GROUP_SIZE_M": 1,
63
+ "num_warps": 8,
64
+ "num_stages": 5
65
+ },
66
+ "64": {
67
+ "BLOCK_SIZE_M": 16,
68
+ "BLOCK_SIZE_N": 128,
69
+ "BLOCK_SIZE_K": 128,
70
+ "GROUP_SIZE_M": 1,
71
+ "num_warps": 8,
72
+ "num_stages": 3
73
+ },
74
+ "96": {
75
+ "BLOCK_SIZE_M": 16,
76
+ "BLOCK_SIZE_N": 64,
77
+ "BLOCK_SIZE_K": 64,
78
+ "GROUP_SIZE_M": 1,
79
+ "num_warps": 4,
80
+ "num_stages": 5
81
+ },
82
+ "128": {
83
+ "BLOCK_SIZE_M": 16,
84
+ "BLOCK_SIZE_N": 128,
85
+ "BLOCK_SIZE_K": 64,
86
+ "GROUP_SIZE_M": 1,
87
+ "num_warps": 4,
88
+ "num_stages": 3
89
+ },
90
+ "256": {
91
+ "BLOCK_SIZE_M": 16,
92
+ "BLOCK_SIZE_N": 64,
93
+ "BLOCK_SIZE_K": 64,
94
+ "GROUP_SIZE_M": 1,
95
+ "num_warps": 4,
96
+ "num_stages": 4
97
+ },
98
+ "512": {
99
+ "BLOCK_SIZE_M": 16,
100
+ "BLOCK_SIZE_N": 256,
101
+ "BLOCK_SIZE_K": 64,
102
+ "GROUP_SIZE_M": 1,
103
+ "num_warps": 4,
104
+ "num_stages": 3
105
+ },
106
+ "1024": {
107
+ "BLOCK_SIZE_M": 64,
108
+ "BLOCK_SIZE_N": 128,
109
+ "BLOCK_SIZE_K": 64,
110
+ "GROUP_SIZE_M": 1,
111
+ "num_warps": 8,
112
+ "num_stages": 3
113
+ },
114
+ "1536": {
115
+ "BLOCK_SIZE_M": 64,
116
+ "BLOCK_SIZE_N": 128,
117
+ "BLOCK_SIZE_K": 64,
118
+ "GROUP_SIZE_M": 1,
119
+ "num_warps": 8,
120
+ "num_stages": 3
121
+ },
122
+ "2048": {
123
+ "BLOCK_SIZE_M": 64,
124
+ "BLOCK_SIZE_N": 64,
125
+ "BLOCK_SIZE_K": 64,
126
+ "GROUP_SIZE_M": 1,
127
+ "num_warps": 4,
128
+ "num_stages": 3
129
+ },
130
+ "3072": {
131
+ "BLOCK_SIZE_M": 64,
132
+ "BLOCK_SIZE_N": 128,
133
+ "BLOCK_SIZE_K": 64,
134
+ "GROUP_SIZE_M": 32,
135
+ "num_warps": 4,
136
+ "num_stages": 3
137
+ },
138
+ "4096": {
139
+ "BLOCK_SIZE_M": 64,
140
+ "BLOCK_SIZE_N": 128,
141
+ "BLOCK_SIZE_K": 64,
142
+ "GROUP_SIZE_M": 32,
143
+ "num_warps": 4,
144
+ "num_stages": 3
145
+ }
146
+ }
sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_4_0/E=512,N=128,device_name=NVIDIA_H100_80GB_HBM3.json ADDED
@@ -0,0 +1,146 @@
1
+ {
2
+ "1": {
3
+ "BLOCK_SIZE_M": 16,
4
+ "BLOCK_SIZE_N": 64,
5
+ "BLOCK_SIZE_K": 64,
6
+ "GROUP_SIZE_M": 1,
7
+ "num_warps": 4,
8
+ "num_stages": 4
9
+ },
10
+ "2": {
11
+ "BLOCK_SIZE_M": 16,
12
+ "BLOCK_SIZE_N": 128,
13
+ "BLOCK_SIZE_K": 64,
14
+ "GROUP_SIZE_M": 1,
15
+ "num_warps": 4,
16
+ "num_stages": 4
17
+ },
18
+ "4": {
19
+ "BLOCK_SIZE_M": 16,
20
+ "BLOCK_SIZE_N": 128,
21
+ "BLOCK_SIZE_K": 64,
22
+ "GROUP_SIZE_M": 1,
23
+ "num_warps": 4,
24
+ "num_stages": 4
25
+ },
26
+ "8": {
27
+ "BLOCK_SIZE_M": 16,
28
+ "BLOCK_SIZE_N": 128,
29
+ "BLOCK_SIZE_K": 64,
30
+ "GROUP_SIZE_M": 1,
31
+ "num_warps": 4,
32
+ "num_stages": 3
33
+ },
34
+ "16": {
35
+ "BLOCK_SIZE_M": 16,
36
+ "BLOCK_SIZE_N": 128,
37
+ "BLOCK_SIZE_K": 64,
38
+ "GROUP_SIZE_M": 1,
39
+ "num_warps": 4,
40
+ "num_stages": 3
41
+ },
42
+ "24": {
43
+ "BLOCK_SIZE_M": 16,
44
+ "BLOCK_SIZE_N": 128,
45
+ "BLOCK_SIZE_K": 64,
46
+ "GROUP_SIZE_M": 1,
47
+ "num_warps": 4,
48
+ "num_stages": 3
49
+ },
50
+ "32": {
51
+ "BLOCK_SIZE_M": 16,
52
+ "BLOCK_SIZE_N": 128,
53
+ "BLOCK_SIZE_K": 64,
54
+ "GROUP_SIZE_M": 1,
55
+ "num_warps": 8,
56
+ "num_stages": 3
57
+ },
58
+ "48": {
59
+ "BLOCK_SIZE_M": 16,
60
+ "BLOCK_SIZE_N": 128,
61
+ "BLOCK_SIZE_K": 128,
62
+ "GROUP_SIZE_M": 1,
63
+ "num_warps": 8,
64
+ "num_stages": 2
65
+ },
66
+ "64": {
67
+ "BLOCK_SIZE_M": 16,
68
+ "BLOCK_SIZE_N": 128,
69
+ "BLOCK_SIZE_K": 128,
70
+ "GROUP_SIZE_M": 1,
71
+ "num_warps": 8,
72
+ "num_stages": 2
73
+ },
74
+ "96": {
75
+ "BLOCK_SIZE_M": 16,
76
+ "BLOCK_SIZE_N": 128,
77
+ "BLOCK_SIZE_K": 128,
78
+ "GROUP_SIZE_M": 1,
79
+ "num_warps": 8,
80
+ "num_stages": 3
81
+ },
82
+ "128": {
83
+ "BLOCK_SIZE_M": 16,
84
+ "BLOCK_SIZE_N": 128,
85
+ "BLOCK_SIZE_K": 64,
86
+ "GROUP_SIZE_M": 1,
87
+ "num_warps": 8,
88
+ "num_stages": 3
89
+ },
90
+ "256": {
91
+ "BLOCK_SIZE_M": 16,
92
+ "BLOCK_SIZE_N": 128,
93
+ "BLOCK_SIZE_K": 64,
94
+ "GROUP_SIZE_M": 1,
95
+ "num_warps": 8,
96
+ "num_stages": 3
97
+ },
98
+ "512": {
99
+ "BLOCK_SIZE_M": 32,
100
+ "BLOCK_SIZE_N": 128,
101
+ "BLOCK_SIZE_K": 64,
102
+ "GROUP_SIZE_M": 1,
103
+ "num_warps": 4,
104
+ "num_stages": 3
105
+ },
106
+ "1024": {
107
+ "BLOCK_SIZE_M": 32,
108
+ "BLOCK_SIZE_N": 128,
109
+ "BLOCK_SIZE_K": 64,
110
+ "GROUP_SIZE_M": 1,
111
+ "num_warps": 4,
112
+ "num_stages": 3
113
+ },
114
+ "1536": {
115
+ "BLOCK_SIZE_M": 64,
116
+ "BLOCK_SIZE_N": 128,
117
+ "BLOCK_SIZE_K": 64,
118
+ "GROUP_SIZE_M": 32,
119
+ "num_warps": 8,
120
+ "num_stages": 3
121
+ },
122
+ "2048": {
123
+ "BLOCK_SIZE_M": 64,
124
+ "BLOCK_SIZE_N": 128,
125
+ "BLOCK_SIZE_K": 64,
126
+ "GROUP_SIZE_M": 16,
127
+ "num_warps": 8,
128
+ "num_stages": 3
129
+ },
130
+ "3072": {
131
+ "BLOCK_SIZE_M": 64,
132
+ "BLOCK_SIZE_N": 128,
133
+ "BLOCK_SIZE_K": 64,
134
+ "GROUP_SIZE_M": 64,
135
+ "num_warps": 8,
136
+ "num_stages": 3
137
+ },
138
+ "4096": {
139
+ "BLOCK_SIZE_M": 128,
140
+ "BLOCK_SIZE_N": 128,
141
+ "BLOCK_SIZE_K": 64,
142
+ "GROUP_SIZE_M": 32,
143
+ "num_warps": 8,
144
+ "num_stages": 3
145
+ }
146
+ }