sglang 0.4.6.post5__py3-none-any.whl → 0.4.7.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.
- sglang/__init__.py +2 -0
- sglang/api.py +7 -0
- sglang/bench_offline_throughput.py +10 -4
- sglang/bench_one_batch_server.py +67 -11
- sglang/bench_serving.py +86 -75
- sglang/lang/backend/runtime_endpoint.py +24 -1
- sglang/lang/interpreter.py +40 -1
- sglang/lang/ir.py +27 -0
- sglang/math_utils.py +8 -0
- sglang/profiler.py +167 -0
- sglang/srt/_custom_ops.py +34 -0
- sglang/srt/configs/internvl.py +8 -12
- sglang/srt/configs/model_config.py +33 -1
- sglang/srt/constrained/base_grammar_backend.py +5 -2
- sglang/srt/constrained/llguidance_backend.py +9 -8
- sglang/srt/constrained/outlines_backend.py +5 -4
- sglang/srt/constrained/xgrammar_backend.py +18 -18
- sglang/srt/conversation.py +52 -8
- sglang/srt/custom_op.py +38 -3
- sglang/srt/debug_utils.py +74 -0
- sglang/srt/disaggregation/base/__init__.py +1 -1
- sglang/srt/disaggregation/base/conn.py +25 -11
- sglang/srt/disaggregation/common/__init__.py +5 -0
- sglang/srt/disaggregation/common/conn.py +407 -0
- sglang/srt/disaggregation/common/utils.py +42 -0
- sglang/srt/disaggregation/decode.py +261 -52
- sglang/srt/disaggregation/fake/__init__.py +1 -1
- sglang/srt/disaggregation/fake/conn.py +16 -9
- sglang/srt/disaggregation/kv_events.py +60 -5
- sglang/srt/disaggregation/launch_lb.py +140 -0
- sglang/srt/disaggregation/mini_lb.py +29 -48
- sglang/srt/disaggregation/mooncake/__init__.py +1 -1
- sglang/srt/disaggregation/mooncake/conn.py +446 -149
- sglang/srt/disaggregation/mooncake/transfer_engine.py +32 -16
- sglang/srt/disaggregation/nixl/__init__.py +6 -1
- sglang/srt/disaggregation/nixl/conn.py +134 -437
- sglang/srt/disaggregation/prefill.py +130 -43
- sglang/srt/disaggregation/utils.py +127 -86
- sglang/srt/distributed/device_communicators/pymscclpp.py +315 -0
- sglang/srt/distributed/parallel_state.py +52 -5
- sglang/srt/entrypoints/EngineBase.py +6 -0
- sglang/srt/entrypoints/engine.py +116 -5
- sglang/srt/entrypoints/http_server.py +28 -4
- sglang/srt/eplb_simulator/__init__.py +1 -0
- sglang/srt/eplb_simulator/reader.py +51 -0
- sglang/srt/function_call/base_format_detector.py +138 -86
- sglang/srt/function_call/deepseekv3_detector.py +54 -6
- sglang/srt/function_call/ebnf_composer.py +33 -19
- sglang/srt/function_call/function_call_parser.py +27 -0
- sglang/srt/function_call/llama32_detector.py +33 -14
- sglang/srt/function_call/mistral_detector.py +73 -26
- sglang/srt/function_call/pythonic_detector.py +86 -20
- sglang/srt/function_call/qwen25_detector.py +64 -10
- sglang/srt/function_call/utils.py +17 -0
- sglang/srt/hf_transformers_utils.py +4 -0
- sglang/srt/layers/activation.py +19 -0
- sglang/srt/layers/attention/aiter_backend.py +503 -125
- sglang/srt/layers/attention/base_attn_backend.py +4 -0
- sglang/srt/layers/attention/cutlass_mla_backend.py +40 -34
- sglang/srt/layers/attention/flashattention_backend.py +137 -63
- sglang/srt/layers/attention/flashinfer_backend.py +46 -3
- sglang/srt/layers/attention/flashinfer_mla_backend.py +59 -25
- sglang/srt/layers/attention/flashmla_backend.py +2 -10
- sglang/srt/layers/attention/intel_amx_backend.py +128 -0
- sglang/srt/layers/attention/tbo_backend.py +232 -0
- sglang/srt/layers/attention/torch_native_backend.py +3 -0
- sglang/srt/layers/attention/triton_backend.py +304 -65
- sglang/srt/layers/attention/triton_ops/decode_attention.py +2 -7
- sglang/srt/layers/attention/triton_ops/extend_attention.py +12 -4
- sglang/srt/layers/attention/vision.py +51 -24
- sglang/srt/layers/communicator.py +281 -197
- sglang/srt/layers/dp_attention.py +6 -5
- sglang/srt/layers/layernorm.py +30 -19
- sglang/srt/layers/linear.py +0 -4
- sglang/srt/layers/logits_processor.py +0 -12
- sglang/srt/layers/moe/cutlass_moe.py +170 -7
- sglang/srt/layers/moe/cutlass_moe_params.py +169 -0
- sglang/srt/layers/moe/ep_moe/kernels.py +33 -11
- sglang/srt/layers/moe/ep_moe/layer.py +136 -72
- sglang/srt/layers/moe/ep_moe/token_dispatcher.py +24 -45
- sglang/srt/layers/moe/fused_moe_native.py +4 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_2_0/E=257,N=128,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_2_0/E=257,N=128,device_name=NVIDIA_H20,dtype=fp8_w8a8,block_shape=[128, 128].json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_2_0/E=257,N=256,device_name=NVIDIA_B200,dtype=fp8_w8a8,block_shape=[128, 128].json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_2_0/E=257,N=256,device_name=NVIDIA_H20,dtype=fp8_w8a8,block_shape=[128, 128].json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_2_0/E=257,N=256,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_2_0/E=8,N=7168,device_name=NVIDIA_H100_80GB_HBM3.json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_3_1/E=257,N=256,device_name=NVIDIA_B200,dtype=fp8_w8a8,block_shape=[128, 128].json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_3_1/E=8,N=7168,device_name=NVIDIA_H100_80GB_HBM3.json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/fused_moe.py +221 -29
- sglang/srt/layers/moe/fused_moe_triton/layer.py +34 -4
- sglang/srt/layers/moe/topk.py +60 -26
- sglang/srt/layers/multimodal.py +3 -3
- sglang/srt/layers/pooler.py +56 -0
- sglang/srt/layers/quantization/__init__.py +3 -2
- sglang/srt/layers/quantization/blockwise_int8.py +3 -0
- sglang/srt/layers/quantization/compressed_tensors/compressed_tensors_moe.py +5 -0
- sglang/srt/layers/quantization/deep_gemm_wrapper/__init__.py +1 -0
- sglang/srt/layers/quantization/{deep_gemm.py → deep_gemm_wrapper/compile_utils.py} +69 -127
- sglang/srt/layers/quantization/deep_gemm_wrapper/configurer.py +32 -0
- sglang/srt/layers/quantization/deep_gemm_wrapper/entrypoint.py +110 -0
- sglang/srt/layers/quantization/fp8.py +28 -23
- sglang/srt/layers/quantization/fp8_kernel.py +156 -75
- sglang/srt/layers/quantization/fp8_utils.py +250 -69
- sglang/srt/layers/quantization/modelopt_quant.py +334 -7
- sglang/srt/layers/quantization/moe_wna16.py +3 -0
- sglang/srt/layers/quantization/w8a8_fp8.py +3 -0
- sglang/srt/layers/quantization/w8a8_int8.py +3 -0
- sglang/srt/layers/radix_attention.py +2 -3
- sglang/srt/layers/rotary_embedding.py +6 -12
- sglang/srt/layers/sampler.py +80 -79
- sglang/srt/layers/utils.py +6 -0
- sglang/srt/lora/layers.py +12 -15
- sglang/srt/lora/lora.py +49 -5
- sglang/srt/lora/lora_manager.py +98 -39
- sglang/srt/lora/mem_pool.py +28 -21
- sglang/srt/lora/utils.py +17 -13
- sglang/srt/managers/cache_controller.py +2 -1
- sglang/srt/managers/data_parallel_controller.py +13 -5
- sglang/srt/managers/eplb_algorithms/__init__.py +63 -0
- sglang/srt/managers/eplb_algorithms/deepseek.py +223 -0
- sglang/srt/managers/{deepseek_eplb.py → eplb_algorithms/deepseek_vec.py} +5 -7
- sglang/srt/managers/eplb_manager.py +55 -14
- sglang/srt/managers/expert_distribution.py +220 -46
- sglang/srt/managers/expert_location.py +110 -56
- sglang/srt/managers/expert_location_dispatch.py +23 -6
- sglang/srt/managers/io_struct.py +43 -8
- sglang/srt/managers/mm_utils.py +88 -38
- sglang/srt/managers/multimodal_processors/base_processor.py +190 -18
- sglang/srt/managers/multimodal_processors/gemma3.py +4 -31
- sglang/srt/managers/multimodal_processors/internvl.py +4 -0
- sglang/srt/managers/multimodal_processors/kimi_vl.py +15 -34
- sglang/srt/managers/multimodal_processors/minicpm.py +2 -1
- sglang/srt/managers/multimodal_processors/phi4mm.py +87 -0
- sglang/srt/managers/multimodal_processors/qwen_vl.py +22 -64
- sglang/srt/managers/multimodal_processors/vila.py +85 -0
- sglang/srt/managers/schedule_batch.py +173 -38
- sglang/srt/managers/scheduler.py +376 -127
- sglang/srt/managers/tokenizer_manager.py +163 -19
- sglang/srt/managers/utils.py +0 -4
- sglang/srt/mem_cache/chunk_cache.py +1 -0
- sglang/srt/mem_cache/hiradix_cache.py +4 -2
- sglang/srt/mem_cache/memory_pool.py +111 -407
- sglang/srt/mem_cache/memory_pool_host.py +380 -0
- sglang/srt/mem_cache/radix_cache.py +36 -12
- sglang/srt/metrics/collector.py +9 -0
- sglang/srt/model_executor/cuda_graph_runner.py +191 -113
- sglang/srt/model_executor/expert_location_updater.py +157 -22
- sglang/srt/model_executor/forward_batch_info.py +52 -22
- sglang/srt/model_executor/model_runner.py +102 -62
- sglang/srt/model_loader/loader.py +8 -1
- sglang/srt/model_loader/utils.py +67 -1
- sglang/srt/models/bert.py +113 -13
- sglang/srt/models/deepseek_nextn.py +1 -1
- sglang/srt/models/deepseek_v2.py +623 -290
- sglang/srt/models/gemma3_causal.py +7 -0
- sglang/srt/models/gemma3_mm.py +19 -14
- sglang/srt/models/idefics2.py +342 -0
- sglang/srt/models/internvl.py +46 -102
- sglang/srt/models/kimi_vl.py +4 -4
- sglang/srt/models/llama.py +1 -1
- sglang/srt/models/minicpmo.py +2 -5
- sglang/srt/models/minicpmv.py +3 -295
- sglang/srt/models/phi4mm.py +512 -0
- sglang/srt/models/qwen2.py +38 -9
- sglang/srt/models/qwen2_5_vl.py +3 -9
- sglang/srt/models/qwen2_eagle.py +4 -1
- sglang/srt/models/qwen2_moe.py +58 -191
- sglang/srt/models/qwen2_vl.py +3 -9
- sglang/srt/models/qwen3.py +41 -10
- sglang/srt/models/qwen3_moe.py +230 -191
- sglang/srt/models/registry.py +9 -1
- sglang/srt/models/roberta.py +117 -9
- sglang/srt/models/transformers.py +291 -0
- sglang/srt/models/vila.py +305 -0
- sglang/srt/openai_api/adapter.py +248 -28
- sglang/srt/openai_api/protocol.py +68 -3
- sglang/srt/openai_api/utils.py +172 -0
- sglang/srt/operations.py +37 -2
- sglang/srt/operations_strategy.py +200 -24
- sglang/srt/sampling/sampling_batch_info.py +37 -1
- sglang/srt/sampling/sampling_params.py +4 -1
- sglang/srt/server_args.py +381 -209
- sglang/srt/speculative/build_eagle_tree.py +9 -9
- sglang/srt/speculative/eagle_draft_cuda_graph_runner.py +12 -14
- sglang/srt/speculative/eagle_draft_extend_cuda_graph_runner.py +256 -0
- sglang/srt/speculative/eagle_utils.py +440 -200
- sglang/srt/speculative/eagle_worker.py +234 -63
- sglang/srt/two_batch_overlap.py +637 -0
- sglang/srt/utils.py +187 -7
- sglang/test/attention/test_prefix_chunk_info.py +2 -0
- sglang/test/runners.py +54 -10
- sglang/test/send_one.py +4 -0
- sglang/test/test_block_fp8.py +1 -0
- sglang/test/test_block_fp8_deep_gemm_blackwell.py +252 -0
- sglang/test/test_block_fp8_ep.py +1 -0
- sglang/test/test_cutlass_moe.py +3 -3
- sglang/test/test_fp4_moe.py +248 -0
- sglang/test/test_utils.py +82 -7
- sglang/utils.py +9 -0
- sglang/version.py +1 -1
- {sglang-0.4.6.post5.dist-info → sglang-0.4.7.post1.dist-info}/METADATA +17 -14
- {sglang-0.4.6.post5.dist-info → sglang-0.4.7.post1.dist-info}/RECORD +359 -321
- {sglang-0.4.6.post5.dist-info → sglang-0.4.7.post1.dist-info}/WHEEL +1 -1
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=1,N=14336,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a16.json → triton_3_1_0/E=1,N=14336,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a16.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=1,N=14336,device_name=NVIDIA_A100-SXM4-80GB.json → triton_3_1_0/E=1,N=14336,device_name=NVIDIA_A100-SXM4-80GB.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=1,N=1792,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a16.json → triton_3_1_0/E=1,N=1792,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a16.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=1,N=1792,device_name=NVIDIA_A100-SXM4-80GB.json → triton_3_1_0/E=1,N=1792,device_name=NVIDIA_A100-SXM4-80GB.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=1,N=3072,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a16.json → triton_3_1_0/E=1,N=3072,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a16.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=1,N=3072,device_name=NVIDIA_H100_80GB_HBM3,dtype=int8_w8a16.json → triton_3_1_0/E=1,N=3072,device_name=NVIDIA_H100_80GB_HBM3,dtype=int8_w8a16.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=1,N=3072,device_name=NVIDIA_H100_80GB_HBM3.json → triton_3_1_0/E=1,N=3072,device_name=NVIDIA_H100_80GB_HBM3.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=1,N=3584,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a16.json → triton_3_1_0/E=1,N=3584,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a16.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=1,N=3584,device_name=NVIDIA_A100-SXM4-80GB.json → triton_3_1_0/E=1,N=3584,device_name=NVIDIA_A100-SXM4-80GB.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=1,N=7168,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a16.json → triton_3_1_0/E=1,N=7168,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a16.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=1,N=7168,device_name=NVIDIA_A100-SXM4-80GB.json → triton_3_1_0/E=1,N=7168,device_name=NVIDIA_A100-SXM4-80GB.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=144,N=512,device_name=NVIDIA_H100_80GB_HBM3.json → triton_3_1_0/E=144,N=512,device_name=NVIDIA_H100_80GB_HBM3.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=16,N=1024,device_name=NVIDIA_H100_80GB_HBM3.json → triton_3_1_0/E=16,N=1024,device_name=NVIDIA_H100_80GB_HBM3.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=16,N=1024,device_name=NVIDIA_H200.json → triton_3_1_0/E=16,N=1024,device_name=NVIDIA_H200.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=16,N=1344,device_name=NVIDIA_A100-SXM4-40GB.json → triton_3_1_0/E=16,N=1344,device_name=NVIDIA_A100-SXM4-40GB.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=16,N=1344,device_name=NVIDIA_A100-SXM4-80GB.json → triton_3_1_0/E=16,N=1344,device_name=NVIDIA_A100-SXM4-80GB.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=16,N=1344,device_name=NVIDIA_H100_80GB_HBM3.json → triton_3_1_0/E=16,N=1344,device_name=NVIDIA_H100_80GB_HBM3.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=16,N=14336,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a16.json → triton_3_1_0/E=16,N=14336,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a16.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=16,N=14336,device_name=NVIDIA_A100-SXM4-80GB.json → triton_3_1_0/E=16,N=14336,device_name=NVIDIA_A100-SXM4-80GB.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=16,N=1792,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a16.json → triton_3_1_0/E=16,N=1792,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a16.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=16,N=1792,device_name=NVIDIA_A100-SXM4-80GB.json → triton_3_1_0/E=16,N=1792,device_name=NVIDIA_A100-SXM4-80GB.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=16,N=2048,device_name=NVIDIA_H100_80GB_HBM3.json → triton_3_1_0/E=16,N=2048,device_name=NVIDIA_H100_80GB_HBM3.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=16,N=2688,device_name=NVIDIA_A100-SXM4-80GB.json → triton_3_1_0/E=16,N=2688,device_name=NVIDIA_A100-SXM4-80GB.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=16,N=2688,device_name=NVIDIA_H100_80GB_HBM3.json → triton_3_1_0/E=16,N=2688,device_name=NVIDIA_H100_80GB_HBM3.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=16,N=3072,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a16.json → triton_3_1_0/E=16,N=3072,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a16.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=16,N=3072,device_name=NVIDIA_H100_80GB_HBM3,dtype=int8_w8a16.json → triton_3_1_0/E=16,N=3072,device_name=NVIDIA_H100_80GB_HBM3,dtype=int8_w8a16.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=16,N=3200,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json → triton_3_1_0/E=16,N=3200,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=16,N=3584,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a16.json → triton_3_1_0/E=16,N=3584,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a16.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=16,N=3584,device_name=NVIDIA_A100-SXM4-80GB.json → triton_3_1_0/E=16,N=3584,device_name=NVIDIA_A100-SXM4-80GB.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=16,N=6400,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json → triton_3_1_0/E=16,N=6400,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=16,N=7168,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a16.json → triton_3_1_0/E=16,N=7168,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a16.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=16,N=7168,device_name=NVIDIA_A100-SXM4-80GB.json → triton_3_1_0/E=16,N=7168,device_name=NVIDIA_A100-SXM4-80GB.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=16,N=7168,device_name=NVIDIA_H100_80GB_HBM3,dtype=int8_w8a16.json → triton_3_1_0/E=16,N=7168,device_name=NVIDIA_H100_80GB_HBM3,dtype=int8_w8a16.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=16,N=800,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json → triton_3_1_0/E=16,N=800,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=160,N=192,device_name=NVIDIA_A800-SXM4-80GB.json → triton_3_1_0/E=160,N=192,device_name=NVIDIA_A800-SXM4-80GB.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=20,N=2048,device_name=NVIDIA_H100_80GB_HBM3.json → triton_3_1_0/E=20,N=2048,device_name=NVIDIA_H100_80GB_HBM3.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=24,N=1024,device_name=NVIDIA_H100_80GB_HBM3.json → triton_3_1_0/E=24,N=1024,device_name=NVIDIA_H100_80GB_HBM3.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=256,N=128,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a8,block_shape=[128, 128].json → triton_3_1_0/E=256,N=128,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a8,block_shape=[128, 128].json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=256,N=128,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a8.json → triton_3_1_0/E=256,N=128,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a8.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=256,N=128,device_name=NVIDIA_A800-SXM4-80GB,dtype=int8_w8a8,block_shape=[128, 128].json → triton_3_1_0/E=256,N=128,device_name=NVIDIA_A800-SXM4-80GB,dtype=int8_w8a8,block_shape=[128, 128].json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=256,N=128,device_name=NVIDIA_A800-SXM4-80GB,dtype=int8_w8a8.json → triton_3_1_0/E=256,N=128,device_name=NVIDIA_A800-SXM4-80GB,dtype=int8_w8a8.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=256,N=128,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json → triton_3_1_0/E=256,N=128,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=256,N=128,device_name=NVIDIA_H20,block_shape=[128, 128].json → triton_3_1_0/E=256,N=128,device_name=NVIDIA_H20,block_shape=[128, 128].json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=256,N=128,device_name=NVIDIA_H20,dtype=fp8_w8a8,block_shape=[128, 128].json → triton_3_1_0/E=256,N=128,device_name=NVIDIA_H20,dtype=fp8_w8a8,block_shape=[128, 128].json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=256,N=128,device_name=NVIDIA_L20Y,dtype=fp8_w8a8,block_shape=[128, 128].json → triton_3_1_0/E=256,N=128,device_name=NVIDIA_L20Y,dtype=fp8_w8a8,block_shape=[128, 128].json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=256,N=256,device_name=AMD_Instinct_MI300X,dtype=fp8_w8a8,block_shape=[128, 128].json → triton_3_1_0/E=256,N=256,device_name=AMD_Instinct_MI300X,dtype=fp8_w8a8,block_shape=[128, 128].json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=256,N=256,device_name=AMD_Instinct_MI325X,dtype=fp8_w8a8,block_shape=[128, 128].json → triton_3_1_0/E=256,N=256,device_name=AMD_Instinct_MI325X,dtype=fp8_w8a8,block_shape=[128, 128].json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=256,N=256,device_name=AMD_Radeon_Graphics,dtype=fp8_w8a8,block_shape=[128, 128].json → triton_3_1_0/E=256,N=256,device_name=AMD_Radeon_Graphics,dtype=fp8_w8a8,block_shape=[128, 128].json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=256,N=256,device_name=NVIDIA_B200,dtype=fp8_w8a8,block_shape=[128, 128].json → triton_3_1_0/E=256,N=256,device_name=NVIDIA_B200,dtype=fp8_w8a8,block_shape=[128, 128].json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=256,N=256,device_name=NVIDIA_H20,dtype=int8_w8a8,block_shape=[128, 128].json → triton_3_1_0/E=256,N=256,device_name=NVIDIA_H20,dtype=int8_w8a8,block_shape=[128, 128].json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=256,N=256,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json → triton_3_1_0/E=256,N=256,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=256,N=64,device_name=NVIDIA_A800-SXM4-80GB.json → triton_3_1_0/E=256,N=64,device_name=NVIDIA_A800-SXM4-80GB.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=256,N=64,device_name=NVIDIA_L20,dtype=int8_w8a8.json → triton_3_1_0/E=256,N=64,device_name=NVIDIA_L20,dtype=int8_w8a8.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=256,N=64,device_name=NVIDIA_L40S,dtype=int8_w8a8.json → triton_3_1_0/E=256,N=64,device_name=NVIDIA_L40S,dtype=int8_w8a8.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=64,N=1024,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json → triton_3_1_0/E=64,N=1024,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=64,N=1280,device_name=NVIDIA_A100-SXM4-80GB.json → triton_3_1_0/E=64,N=1280,device_name=NVIDIA_A100-SXM4-80GB.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=64,N=1280,device_name=NVIDIA_A800-SXM4-80GB.json → triton_3_1_0/E=64,N=1280,device_name=NVIDIA_A800-SXM4-80GB.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=64,N=1280,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json → triton_3_1_0/E=64,N=1280,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=64,N=1280,device_name=NVIDIA_H100_80GB_HBM3.json → triton_3_1_0/E=64,N=1280,device_name=NVIDIA_H100_80GB_HBM3.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=64,N=1280,device_name=NVIDIA_H200,dtype=fp8_w8a8.json → triton_3_1_0/E=64,N=1280,device_name=NVIDIA_H200,dtype=fp8_w8a8.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=64,N=1280,device_name=NVIDIA_H200.json → triton_3_1_0/E=64,N=1280,device_name=NVIDIA_H200.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=64,N=2560,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json → triton_3_1_0/E=64,N=2560,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=64,N=2560,device_name=NVIDIA_H200,dtype=fp8_w8a8.json → triton_3_1_0/E=64,N=2560,device_name=NVIDIA_H200,dtype=fp8_w8a8.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=64,N=2560,device_name=NVIDIA_H200.json → triton_3_1_0/E=64,N=2560,device_name=NVIDIA_H200.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=64,N=320,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json → triton_3_1_0/E=64,N=320,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=64,N=320,device_name=NVIDIA_H100_80GB_HBM3.json → triton_3_1_0/E=64,N=320,device_name=NVIDIA_H100_80GB_HBM3.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=64,N=320,device_name=NVIDIA_H200,dtype=fp8_w8a8.json → triton_3_1_0/E=64,N=320,device_name=NVIDIA_H200,dtype=fp8_w8a8.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=64,N=320,device_name=NVIDIA_H200.json → triton_3_1_0/E=64,N=320,device_name=NVIDIA_H200.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=64,N=512,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json → triton_3_1_0/E=64,N=512,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=64,N=640,device_name=NVIDIA_A100-SXM4-80GB.json → triton_3_1_0/E=64,N=640,device_name=NVIDIA_A100-SXM4-80GB.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=64,N=640,device_name=NVIDIA_A800-SXM4-80GB.json → triton_3_1_0/E=64,N=640,device_name=NVIDIA_A800-SXM4-80GB.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=64,N=640,device_name=NVIDIA_GeForce_RTX_4090,dtype=fp8_w8a8.json → triton_3_1_0/E=64,N=640,device_name=NVIDIA_GeForce_RTX_4090,dtype=fp8_w8a8.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=64,N=640,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json → triton_3_1_0/E=64,N=640,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=64,N=640,device_name=NVIDIA_H100_80GB_HBM3.json → triton_3_1_0/E=64,N=640,device_name=NVIDIA_H100_80GB_HBM3.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=64,N=640,device_name=NVIDIA_H200,dtype=fp8_w8a8.json → triton_3_1_0/E=64,N=640,device_name=NVIDIA_H200,dtype=fp8_w8a8.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=64,N=640,device_name=NVIDIA_H200.json → triton_3_1_0/E=64,N=640,device_name=NVIDIA_H200.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=14336,device_name=AMD_Instinct_MI300X.json → triton_3_1_0/E=8,N=14336,device_name=AMD_Instinct_MI300X.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=14336,device_name=AMD_Instinct_MI325X.json → triton_3_1_0/E=8,N=14336,device_name=AMD_Instinct_MI325X.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=14336,device_name=AMD_Radeon_Graphics.json → triton_3_1_0/E=8,N=14336,device_name=AMD_Radeon_Graphics.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=14336,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json → triton_3_1_0/E=8,N=14336,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=14336,device_name=NVIDIA_H200,dtype=fp8_w8a8.json → triton_3_1_0/E=8,N=14336,device_name=NVIDIA_H200,dtype=fp8_w8a8.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=14336,device_name=NVIDIA_H200.json → triton_3_1_0/E=8,N=14336,device_name=NVIDIA_H200.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=1792,device_name=AMD_Instinct_MI300X.json → triton_3_1_0/E=8,N=1792,device_name=AMD_Instinct_MI300X.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=1792,device_name=AMD_Instinct_MI325X.json → triton_3_1_0/E=8,N=1792,device_name=AMD_Instinct_MI325X.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=1792,device_name=AMD_Radeon_Graphics.json → triton_3_1_0/E=8,N=1792,device_name=AMD_Radeon_Graphics.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=1792,device_name=NVIDIA_A100-SXM4-40GB.json → triton_3_1_0/E=8,N=1792,device_name=NVIDIA_A100-SXM4-40GB.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=1792,device_name=NVIDIA_A100-SXM4-80GB.json → triton_3_1_0/E=8,N=1792,device_name=NVIDIA_A100-SXM4-80GB.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=1792,device_name=NVIDIA_H100_80GB_HBM3.json → triton_3_1_0/E=8,N=1792,device_name=NVIDIA_H100_80GB_HBM3.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=1792,device_name=NVIDIA_H200,dtype=fp8_w8a8.json → triton_3_1_0/E=8,N=1792,device_name=NVIDIA_H200,dtype=fp8_w8a8.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=1792,device_name=NVIDIA_H200.json → triton_3_1_0/E=8,N=1792,device_name=NVIDIA_H200.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=2048,device_name=NVIDIA_A100-SXM4-80GB.json → triton_3_1_0/E=8,N=2048,device_name=NVIDIA_A100-SXM4-80GB.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=2048,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json → triton_3_1_0/E=8,N=2048,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=2048,device_name=NVIDIA_H100_80GB_HBM3.json → triton_3_1_0/E=8,N=2048,device_name=NVIDIA_H100_80GB_HBM3.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=2048,device_name=NVIDIA_H200,dtype=fp8_w8a8.json → triton_3_1_0/E=8,N=2048,device_name=NVIDIA_H200,dtype=fp8_w8a8.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=2048,device_name=NVIDIA_H200.json → triton_3_1_0/E=8,N=2048,device_name=NVIDIA_H200.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=3584,device_name=AMD_Instinct_MI300X.json → triton_3_1_0/E=8,N=3584,device_name=AMD_Instinct_MI300X.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=3584,device_name=AMD_Instinct_MI325X.json → triton_3_1_0/E=8,N=3584,device_name=AMD_Instinct_MI325X.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=3584,device_name=AMD_Radeon_Graphics.json → triton_3_1_0/E=8,N=3584,device_name=AMD_Radeon_Graphics.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=3584,device_name=NVIDIA_A100-SXM4-40GB.json → triton_3_1_0/E=8,N=3584,device_name=NVIDIA_A100-SXM4-40GB.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=3584,device_name=NVIDIA_A100-SXM4-80GB.json → triton_3_1_0/E=8,N=3584,device_name=NVIDIA_A100-SXM4-80GB.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=3584,device_name=NVIDIA_GeForce_RTX_4090,dtype=fp8_w8a8.json → triton_3_1_0/E=8,N=3584,device_name=NVIDIA_GeForce_RTX_4090,dtype=fp8_w8a8.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=3584,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json → triton_3_1_0/E=8,N=3584,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=3584,device_name=NVIDIA_H100_80GB_HBM3.json → triton_3_1_0/E=8,N=3584,device_name=NVIDIA_H100_80GB_HBM3.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=3584,device_name=NVIDIA_H200,dtype=fp8_w8a8.json → triton_3_1_0/E=8,N=3584,device_name=NVIDIA_H200,dtype=fp8_w8a8.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=3584,device_name=NVIDIA_H200.json → triton_3_1_0/E=8,N=3584,device_name=NVIDIA_H200.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=3584,device_name=NVIDIA_L40S.json → triton_3_1_0/E=8,N=3584,device_name=NVIDIA_L40S.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=4096,device_name=AMD_Instinct_MI300X,dtype=fp8_w8a8.json → triton_3_1_0/E=8,N=4096,device_name=AMD_Instinct_MI300X,dtype=fp8_w8a8.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=4096,device_name=AMD_Instinct_MI325X,dtype=fp8_w8a8.json → triton_3_1_0/E=8,N=4096,device_name=AMD_Instinct_MI325X,dtype=fp8_w8a8.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=4096,device_name=AMD_Radeon_Graphics,dtype=fp8_w8a8.json → triton_3_1_0/E=8,N=4096,device_name=AMD_Radeon_Graphics,dtype=fp8_w8a8.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=4096,device_name=NVIDIA_A100-SXM4-80GB.json → triton_3_1_0/E=8,N=4096,device_name=NVIDIA_A100-SXM4-80GB.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=4096,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json → triton_3_1_0/E=8,N=4096,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=4096,device_name=NVIDIA_H100_80GB_HBM3.json → triton_3_1_0/E=8,N=4096,device_name=NVIDIA_H100_80GB_HBM3.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=4096,device_name=NVIDIA_H200,dtype=fp8_w8a8.json → triton_3_1_0/E=8,N=4096,device_name=NVIDIA_H200,dtype=fp8_w8a8.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=4096,device_name=NVIDIA_H200.json → triton_3_1_0/E=8,N=4096,device_name=NVIDIA_H200.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=7168,device_name=AMD_Instinct_MI300X.json → triton_3_1_0/E=8,N=7168,device_name=AMD_Instinct_MI300X.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=7168,device_name=AMD_Instinct_MI325X.json → triton_3_1_0/E=8,N=7168,device_name=AMD_Instinct_MI325X.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=7168,device_name=AMD_Radeon_Graphics.json → triton_3_1_0/E=8,N=7168,device_name=AMD_Radeon_Graphics.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=7168,device_name=NVIDIA_A100-SXM4-80GB.json → triton_3_1_0/E=8,N=7168,device_name=NVIDIA_A100-SXM4-80GB.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=7168,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json → triton_3_1_0/E=8,N=7168,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=7168,device_name=NVIDIA_H100_80GB_HBM3.json → triton_3_1_0/E=8,N=7168,device_name=NVIDIA_H100_80GB_HBM3.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=7168,device_name=NVIDIA_H200,dtype=fp8_w8a8.json → triton_3_1_0/E=8,N=7168,device_name=NVIDIA_H200,dtype=fp8_w8a8.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=7168,device_name=NVIDIA_H200.json → triton_3_1_0/E=8,N=7168,device_name=NVIDIA_H200.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=8192,device_name=AMD_Instinct_MI300X,dtype=fp8_w8a8.json → triton_3_1_0/E=8,N=8192,device_name=AMD_Instinct_MI300X,dtype=fp8_w8a8.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=8192,device_name=AMD_Instinct_MI325X,dtype=fp8_w8a8.json → triton_3_1_0/E=8,N=8192,device_name=AMD_Instinct_MI325X,dtype=fp8_w8a8.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=8192,device_name=AMD_Radeon_Graphics,dtype=fp8_w8a8.json → triton_3_1_0/E=8,N=8192,device_name=AMD_Radeon_Graphics,dtype=fp8_w8a8.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=8192,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json → triton_3_1_0/E=8,N=8192,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=8,N=8192,device_name=NVIDIA_H200,dtype=fp8_w8a8.json → triton_3_1_0/E=8,N=8192,device_name=NVIDIA_H200,dtype=fp8_w8a8.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=128,N=192,device_name=NVIDIA_A800-SXM4-80GB.json → triton_3_2_0/E=128,N=192,device_name=NVIDIA_A800-SXM4-80GB.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=128,N=192,device_name=NVIDIA_H100_80GB_HBM3.json → triton_3_2_0/E=128,N=192,device_name=NVIDIA_H100_80GB_HBM3.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=128,N=192,device_name=NVIDIA_H20.json → triton_3_2_0/E=128,N=192,device_name=NVIDIA_H20.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=128,N=192,device_name=NVIDIA_H200.json → triton_3_2_0/E=128,N=192,device_name=NVIDIA_H200.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=128,N=384,device_name=NVIDIA_H100_80GB_HBM3.json → triton_3_2_0/E=128,N=384,device_name=NVIDIA_H100_80GB_HBM3.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=128,N=384,device_name=NVIDIA_H20,dtype=fp8_w8a8,block_shape=[128, 128].json → triton_3_2_0/E=128,N=384,device_name=NVIDIA_H20,dtype=fp8_w8a8,block_shape=[128, 128].json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=128,N=384,device_name=NVIDIA_H20.json → triton_3_2_0/E=128,N=384,device_name=NVIDIA_H20.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=128,N=384,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json → triton_3_2_0/E=128,N=384,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=128,N=384,device_name=NVIDIA_H200.json → triton_3_2_0/E=128,N=384,device_name=NVIDIA_H200.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=128,N=512,device_name=NVIDIA_H100_80GB_HBM3.json → triton_3_2_0/E=128,N=512,device_name=NVIDIA_H100_80GB_HBM3.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=128,N=768,device_name=NVIDIA_A800-SXM4-80GB.json → triton_3_2_0/E=128,N=768,device_name=NVIDIA_A800-SXM4-80GB.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=128,N=768,device_name=NVIDIA_H100_80GB_HBM3.json → triton_3_2_0/E=128,N=768,device_name=NVIDIA_H100_80GB_HBM3.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=128,N=768,device_name=NVIDIA_H20,dtype=fp8_w8a8,block_shape=[128, 128].json → triton_3_2_0/E=128,N=768,device_name=NVIDIA_H20,dtype=fp8_w8a8,block_shape=[128, 128].json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=128,N=768,device_name=NVIDIA_H20.json → triton_3_2_0/E=128,N=768,device_name=NVIDIA_H20.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=128,N=768,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json → triton_3_2_0/E=128,N=768,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=128,N=768,device_name=NVIDIA_H200.json → triton_3_2_0/E=128,N=768,device_name=NVIDIA_H200.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=128,N=96,device_name=NVIDIA_H20.json → triton_3_2_0/E=128,N=96,device_name=NVIDIA_H20.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=264,N=128,device_name=NVIDIA_A800-SXM4-80GB,dtype=int8_w8a8.json → triton_3_2_0/E=264,N=128,device_name=NVIDIA_A800-SXM4-80GB,dtype=int8_w8a8.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=264,N=256,device_name=NVIDIA_B200,dtype=fp8_w8a8,block_shape=[128, 128].json → triton_3_2_0/E=264,N=256,device_name=NVIDIA_B200,dtype=fp8_w8a8,block_shape=[128, 128].json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=264,N=256,device_name=NVIDIA_H20,dtype=fp8_w8a8,block_shape=[128, 128].json → triton_3_2_0/E=264,N=256,device_name=NVIDIA_H20,dtype=fp8_w8a8,block_shape=[128, 128].json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=264,N=256,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json → triton_3_2_0/E=264,N=256,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=272,N=128,device_name=NVIDIA_A800-SXM4-80GB,dtype=int8_w8a8.json → triton_3_2_0/E=272,N=128,device_name=NVIDIA_A800-SXM4-80GB,dtype=int8_w8a8.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=272,N=128,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json → triton_3_2_0/E=272,N=128,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=272,N=128,device_name=NVIDIA_H20,dtype=fp8_w8a8,block_shape=[128, 128].json → triton_3_2_0/E=272,N=128,device_name=NVIDIA_H20,dtype=fp8_w8a8,block_shape=[128, 128].json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=272,N=64,device_name=NVIDIA_A800-SXM4-80GB.json → triton_3_2_0/E=272,N=64,device_name=NVIDIA_A800-SXM4-80GB.json} +0 -0
- /sglang/srt/layers/moe/fused_moe_triton/configs/{E=288,N=64,device_name=NVIDIA_A800-SXM4-80GB.json → triton_3_2_0/E=288,N=64,device_name=NVIDIA_A800-SXM4-80GB.json} +0 -0
- {sglang-0.4.6.post5.dist-info → sglang-0.4.7.post1.dist-info}/licenses/LICENSE +0 -0
- {sglang-0.4.6.post5.dist-info → sglang-0.4.7.post1.dist-info}/top_level.txt +0 -0
@@ -30,16 +30,24 @@ class MooncakeTransferEngine:
|
|
30
30
|
self.session_id = f"{self.hostname}:{self.engine.get_rpc_port()}"
|
31
31
|
|
32
32
|
def register(self, ptr, length):
|
33
|
-
|
33
|
+
try:
|
34
|
+
ret_value = self.engine.register_memory(ptr, length)
|
35
|
+
except Exception:
|
36
|
+
# Mark register as failed
|
37
|
+
ret_value = -1
|
38
|
+
|
34
39
|
if ret_value != 0:
|
35
|
-
logger.
|
36
|
-
raise RuntimeError("Mooncake memory registration failed.")
|
40
|
+
logger.debug("Mooncake memory registration %s failed.", ptr)
|
37
41
|
|
38
42
|
def deregister(self, ptr):
|
39
|
-
|
43
|
+
try:
|
44
|
+
ret_value = self.engine.unregister_memory(ptr)
|
45
|
+
except Exception:
|
46
|
+
# Mark deregister as failed
|
47
|
+
ret_value = -1
|
48
|
+
|
40
49
|
if ret_value != 0:
|
41
|
-
logger.
|
42
|
-
raise RuntimeError("Mooncake memory deregistration failed.")
|
50
|
+
logger.debug("Mooncake memory deregistration %s failed.", ptr)
|
43
51
|
|
44
52
|
def initialize(
|
45
53
|
self,
|
@@ -61,18 +69,26 @@ class MooncakeTransferEngine:
|
|
61
69
|
self, session_id: str, buffer: int, peer_buffer_address: int, length: int
|
62
70
|
) -> int:
|
63
71
|
"""Synchronously transfer data to the specified address."""
|
64
|
-
|
65
|
-
|
66
|
-
|
67
|
-
|
68
|
-
|
72
|
+
try:
|
73
|
+
# the first time: based on session_id (which contains remote_ip) to construct a queue pair, and cache the queue pair
|
74
|
+
# later: based on the cached queue pair to send data
|
75
|
+
ret = self.engine.transfer_sync_write(
|
76
|
+
session_id, buffer, peer_buffer_address, length
|
77
|
+
)
|
78
|
+
except Exception:
|
79
|
+
# Mark transfer request as failed
|
80
|
+
ret = -1
|
81
|
+
|
69
82
|
if ret < 0:
|
70
|
-
|
71
|
-
|
72
|
-
|
83
|
+
# Do not raise an exception here, since some transfer requests fail should be accepted and the execution thread should not be stopped.
|
84
|
+
logger.debug(
|
85
|
+
"Failed to transfer data from %s to %s - %s.",
|
86
|
+
buffer,
|
87
|
+
session_id,
|
88
|
+
peer_buffer_address,
|
89
|
+
)
|
73
90
|
|
74
|
-
|
75
|
-
return self.hostname
|
91
|
+
return ret
|
76
92
|
|
77
93
|
def get_session_id(self):
|
78
94
|
return self.session_id
|