chore: upgrade sgl-kernel v0.1.2.post1 (sgl-project#6196)

Co-authored-by: alcanderian <[email protected]>

Author: 2 people

python/pyproject.toml

@@ -48,7 +48,7 @@ runtime_common = [
 
 srt = [
     "sglang[runtime_common]",
-    "sgl-kernel==0.1.1",
+    "sgl-kernel==0.1.2.post1",
     "flashinfer_python==0.2.5",
     "torch==2.6.0",
     "torchvision==0.21.0",

python/sglang/srt/entrypoints/engine.py

@@ -486,7 +486,7 @@ def _set_envs_and_config(server_args: ServerArgs):
     if _is_cuda:
         assert_pkg_version(
             "sgl-kernel",
-            "0.1.1",
+            "0.1.2.post1",
             "Please reinstall the latest version with `pip install sgl-kernel --force-reinstall`",
         )
 

python/sglang/srt/layers/quantization/deep_gemm.py

@@ -16,11 +16,7 @@
     import deep_gemm
     from deep_gemm import get_num_sms
     from deep_gemm.jit_kernels.gemm import get_best_configs
-    from deep_gemm.jit_kernels.gemm import includes as deep_gemm_includes
-    from deep_gemm.jit_kernels.gemm import template as deep_gemm_gemm_template
-    from deep_gemm.jit_kernels.m_grouped_gemm import (
-        template as deep_gemm_grouped_gemm_template,
-    )
+    from deep_gemm.jit_kernels.runtime import FP8GemmRuntime, GemmType
     from deep_gemm.jit_kernels.tuner import jit_tuner
 
     sm_version = get_device_sm()
@@ -45,10 +41,15 @@ def get_enable_jit_deepgemm():
 _IN_PRECOMPILE_STAGE = get_bool_env_var("SGL_IN_DEEPGEMM_PRECOMPILE_STAGE", "false")
 
 # Force redirect deep_gemm cache_dir
-os.environ["DG_CACHE_DIR"] = os.getenv(
-    "SGL_DG_CACHE_DIR", os.path.expanduser("~") + "/.cache/deep_gemm"
+os.environ["DG_JIT_CACHE_DIR"] = os.getenv(
+    "SGL_DG_CACHE_DIR", os.path.join(os.path.expanduser("~"), ".cache", "deep_gemm")
 )
 
+# Refer to https://github.com/deepseek-ai/DeepGEMM/commit/d75b218b7b8f4a5dd5406ac87905039ead3ae42f
+# NVRTC may have performance loss with some cases.
+# And NVCC JIT speed is also 9x faster in the ref commit
+os.environ["DG_JIT_USE_NVRTC"] = os.getenv("SGL_DG_USE_NVRTC", "0")
+
 
 def update_deep_gemm_config(gpu_id: int, server_args: ServerArgs):
     global _BUILTIN_M_LIST
@@ -130,10 +131,18 @@ def _compile_grouped_gemm_nt_f8f8bf16_masked_one(
     num_groups: int,
     config: Tuple[int, int, int, int, Tuple[int, bool], Tuple[int, int, int]],
 ) -> None:
-    # Auto-tuning with compilation
-    global deep_gemm_includes, deep_gemm_grouped_gemm_template
-    _, block_m, block_n, num_stages, tma_multicast_config, smem_config = config
-    _ = jit_tuner.compile_and_tune(
+    num_sms, block_m, block_n, num_stages, tma_multicast_config, smem_config = config
+    block_k = 128
+    num_tma_threads = 128
+    num_math_threads_per_group = 128
+    kwargs = {
+        "NUM_TMA_THREADS": num_tma_threads,
+        "NUM_MATH_THREADS_PER_GROUP": num_math_threads_per_group,
+        "BLOCK_K": block_k,
+        "NUM_SMS": num_sms,
+        "SMEM_SIZE": smem_config[0],
+    }
+    _, _ = jit_tuner.compile_and_tune(
         name="m_grouped_gemm_fp8_fp8_bf16_nt",
         keys={
             "N": n,
@@ -146,24 +155,11 @@ def _compile_grouped_gemm_nt_f8f8bf16_masked_one(
             "NUM_STAGES": num_stages,
             "NUM_TMA_MULTICAST": tma_multicast_config[0],
             "IS_TMA_MULTICAST_ON_A": tma_multicast_config[1],
-            "GEMM_TYPE": "GroupedMasked",
+            "GEMM_TYPE": GemmType.GroupedMasked,
         },
         space=(),
-        includes=deep_gemm_includes,
-        arg_defs=(
-            ("lhs", torch.float8_e4m3fn),
-            ("lhs_scales", torch.float),
-            ("rhs", torch.float8_e4m3fn),
-            ("rhs_scales", torch.float),
-            ("out", torch.bfloat16),
-            ("grouped_layout", torch.int32),
-            ("m", int),
-            ("stream", torch.cuda.Stream),
-            ("num_sms", int),
-            ("smem_size", int),
-        ),
-        template=deep_gemm_grouped_gemm_template,
-        args=[],
+        kwargs=kwargs,
+        runtime_cls=FP8GemmRuntime,
     )
 
 
@@ -173,9 +169,18 @@ def _compile_grouped_gemm_nt_f8f8bf16_contig_one(
     num_groups: int,
     config: Tuple[int, int, int, int, Tuple[int, bool], Tuple[int, int, int]],
 ) -> None:
-    global deep_gemm_includes, deep_gemm_grouped_gemm_template
-    _, block_m, block_n, num_stages, tma_multicast_config, smem_config = config
-    _ = jit_tuner.compile_and_tune(
+    num_sms, block_m, block_n, num_stages, tma_multicast_config, smem_config = config
+    block_k = 128
+    num_tma_threads = 128
+    num_math_threads_per_group = 128
+    kwargs = {
+        "NUM_TMA_THREADS": num_tma_threads,
+        "NUM_MATH_THREADS_PER_GROUP": num_math_threads_per_group,
+        "BLOCK_K": block_k,
+        "NUM_SMS": num_sms,
+        "SMEM_SIZE": smem_config[0],
+    }
+    _, _ = jit_tuner.compile_and_tune(
         name="m_grouped_gemm_fp8_fp8_bf16_nt",
         keys={
             "N": n,
@@ -188,25 +193,11 @@ def _compile_grouped_gemm_nt_f8f8bf16_contig_one(
             "NUM_STAGES": num_stages,
             "NUM_TMA_MULTICAST": tma_multicast_config[0],
             "IS_TMA_MULTICAST_ON_A": tma_multicast_config[1],
-            "GEMM_TYPE": "GroupedContiguous",
+            "GEMM_TYPE": GemmType.GroupedContiguous,
         },
         space=(),
-        includes=deep_gemm_includes,
-        arg_defs=(
-            ("lhs", torch.float8_e4m3fn),
-            ("lhs_scales", torch.float),
-            ("rhs", torch.float8_e4m3fn),
-            ("rhs_scales", torch.float),
-            ("out", torch.bfloat16),
-            ("grouped_layout", torch.int32),
-            ("m", int),
-            ("num_groups", int),
-            ("stream", torch.cuda.Stream),
-            ("num_sms", int),
-            ("smem_size", int),
-        ),
-        template=deep_gemm_grouped_gemm_template,
-        args=[],
+        kwargs=kwargs,
+        runtime_cls=FP8GemmRuntime,
     )
 
 
@@ -216,9 +207,20 @@ def _compile_gemm_nt_f8f8bf16_one(
     _: int,  # _ is a dummy parameter to align with other interfaces
     config: Tuple[int, int, int, int, Tuple[int, bool], Tuple[int, int, int]],
 ) -> None:
-    global deep_gemm_includes, deep_gemm_gemm_template
-    _, block_m, block_n, num_stages, tma_multicast_config, smem_config = config
-    _ = jit_tuner.compile_and_tune(
+    num_sms, block_m, block_n, num_stages, tma_multicast_config, smem_config = config
+    block_k = 128
+    num_tma_threads = 128
+    num_math_threads_per_group = 128
+    kwargs = {
+        "GEMM_TYPE": GemmType.Normal,
+        "NUM_TMA_THREADS": num_tma_threads,
+        "NUM_MATH_THREADS_PER_GROUP": num_math_threads_per_group,
+        "NUM_GROUPS": 1,
+        "BLOCK_K": block_k,
+        "NUM_SMS": num_sms,
+        "SMEM_SIZE": smem_config[0],
+    }
+    _, _ = jit_tuner.compile_and_tune(
         name="gemm_fp8_fp8_bf16_nt",
         keys={
             "N": n,
@@ -232,20 +234,8 @@ def _compile_gemm_nt_f8f8bf16_one(
             "IS_TMA_MULTICAST_ON_A": tma_multicast_config[1],
         },
         space=(),
-        includes=deep_gemm_includes,
-        arg_defs=(
-            ("lhs", torch.float8_e4m3fn),
-            ("lhs_scales", torch.float),
-            ("rhs", torch.float8_e4m3fn),
-            ("rhs_scales", torch.float),
-            ("out", torch.bfloat16),
-            ("m", int),
-            ("stream", torch.cuda.Stream),
-            ("num_sms", int),
-            ("smem_size", int),
-        ),
-        template=deep_gemm_gemm_template,
-        args=[],
+        kwargs=kwargs,
+        runtime_cls=FP8GemmRuntime,
     )
 
 
@@ -373,7 +363,7 @@ def _log_jit_build(M: int, N: int, K: int, kernel_type: DeepGemmKernelType):
 
     from deep_gemm.jit.runtime import RuntimeCache
 
-    origin_func = RuntimeCache.__getitem__
+    origin_func = RuntimeCache.get
 
     def __patched_func(self, *args, **kwargs):
         ret = origin_func(self, *args, **kwargs)
@@ -385,6 +375,6 @@ def __patched_func(self, *args, **kwargs):
             )
         return ret
 
-    RuntimeCache.__getitem__ = __patched_func
+    RuntimeCache.get = __patched_func
     yield
-    RuntimeCache.__getitem__ = origin_func
+    RuntimeCache.get = origin_func

scripts/ci_install_dependency.sh

@@ -16,7 +16,7 @@ rm -rf /usr/local/lib/python3.10/dist-packages/sgl_kernel*
 pip install --upgrade pip
 
 # Install sgl-kernel
-pip install sgl-kernel==0.1.1 --no-cache-dir
+pip install sgl-kernel==0.1.2.post1 --no-cache-dir
 
 # Install the main package
 pip install -e "python[all]"

scripts/ci_install_dependency_8_gpu.sh

@@ -34,7 +34,7 @@ rm -rf /usr/local/include/nvshmem*
 pip install --upgrade pip
 
 # Install sgl-kernel
-pip install sgl-kernel==0.1.1 --no-cache-dir
+pip install sgl-kernel==0.1.2.post1 --no-cache-dir
 
 # Install the main package
 pip install -e "python[all]"