[FlashInfer] Bump FlashInfer version from 0.6.2 to 0.6.3 (#18448)
This commit is contained in:
@@ -10,9 +10,9 @@ on:
|
|||||||
description: "Version to build (without v prefix, e.g., 0.5.8)"
|
description: "Version to build (without v prefix, e.g., 0.5.8)"
|
||||||
required: true
|
required: true
|
||||||
flashinfer_version:
|
flashinfer_version:
|
||||||
description: "FlashInfer version (default: 0.6.1)"
|
description: "FlashInfer version (default: 0.6.3)"
|
||||||
required: false
|
required: false
|
||||||
default: "0.6.1"
|
default: "0.6.3"
|
||||||
|
|
||||||
jobs:
|
jobs:
|
||||||
publish-x86:
|
publish-x86:
|
||||||
|
|||||||
+1
-1
@@ -19,7 +19,7 @@ ARG PIP_DEFAULT_INDEX
|
|||||||
ARG UBUNTU_MIRROR
|
ARG UBUNTU_MIRROR
|
||||||
ARG GITHUB_ARTIFACTORY=github.com
|
ARG GITHUB_ARTIFACTORY=github.com
|
||||||
ARG INSTALL_FLASHINFER_JIT_CACHE=0
|
ARG INSTALL_FLASHINFER_JIT_CACHE=0
|
||||||
ARG FLASHINFER_VERSION=0.6.2
|
ARG FLASHINFER_VERSION=0.6.3
|
||||||
ARG MOONCAKE_VERSION=0.3.9
|
ARG MOONCAKE_VERSION=0.3.9
|
||||||
#if need other arg please add in MOONCAKE_COMPILE_ARG
|
#if need other arg please add in MOONCAKE_COMPILE_ARG
|
||||||
ARG MOONCAKE_COMPILE_ARG="-DUSE_HTTP=ON -DUSE_MNNVL=ON -DUSE_CUDA=ON -DWITH_EP=ON"
|
ARG MOONCAKE_COMPILE_ARG="-DUSE_HTTP=ON -DUSE_MNNVL=ON -DUSE_CUDA=ON -DWITH_EP=ON"
|
||||||
|
|||||||
@@ -28,8 +28,8 @@ dependencies = [
|
|||||||
"datasets",
|
"datasets",
|
||||||
"einops",
|
"einops",
|
||||||
"fastapi",
|
"fastapi",
|
||||||
"flashinfer_python==0.6.2", # keep it aligned with jit-cache version in Dockerfile
|
"flashinfer_python==0.6.3", # keep it aligned with jit-cache version in Dockerfile
|
||||||
"flashinfer_cubin==0.6.2",
|
"flashinfer_cubin==0.6.3",
|
||||||
"gguf",
|
"gguf",
|
||||||
"hf_transfer",
|
"hf_transfer",
|
||||||
"huggingface_hub",
|
"huggingface_hub",
|
||||||
|
|||||||
@@ -841,7 +841,7 @@ def _set_envs_and_config(server_args: ServerArgs):
|
|||||||
if server_args.attention_backend == "flashinfer":
|
if server_args.attention_backend == "flashinfer":
|
||||||
assert_pkg_version(
|
assert_pkg_version(
|
||||||
"flashinfer_python",
|
"flashinfer_python",
|
||||||
"0.6.2",
|
"0.6.3",
|
||||||
"Please uninstall the old version and "
|
"Please uninstall the old version and "
|
||||||
"reinstall the latest version by following the instructions "
|
"reinstall the latest version by following the instructions "
|
||||||
"at https://docs.flashinfer.ai/installation.html.",
|
"at https://docs.flashinfer.ai/installation.html.",
|
||||||
|
|||||||
@@ -1616,8 +1616,8 @@ class ServerArgs:
|
|||||||
and self.moe_a2a_backend == "none"
|
and self.moe_a2a_backend == "none"
|
||||||
and self.moe_runner_backend == "auto"
|
and self.moe_runner_backend == "auto"
|
||||||
):
|
):
|
||||||
# Only enable flashinfer_trtllm if flashinfer-python version is >= 0.6.2
|
# Only enable flashinfer_trtllm if flashinfer-python version is >= 0.6.3
|
||||||
if check_pkg_version_at_least("flashinfer-python", "0.6.2"):
|
if check_pkg_version_at_least("flashinfer-python", "0.6.3"):
|
||||||
self.moe_runner_backend = "flashinfer_trtllm"
|
self.moe_runner_backend = "flashinfer_trtllm"
|
||||||
logger.info(
|
logger.info(
|
||||||
"Use flashinfer_trtllm as MoE runner backend on sm100 for Glm4MoeForCausalLM"
|
"Use flashinfer_trtllm as MoE runner backend on sm100 for Glm4MoeForCausalLM"
|
||||||
|
|||||||
@@ -1113,7 +1113,7 @@ def check_pkg_version_at_least(pkg: str, min_version: str) -> bool:
|
|||||||
|
|
||||||
Args:
|
Args:
|
||||||
pkg: Package name (distribution name, e.g., "flashinfer-python")
|
pkg: Package name (distribution name, e.g., "flashinfer-python")
|
||||||
min_version: Minimum version required (e.g., "0.6.2")
|
min_version: Minimum version required (e.g., "0.6.3")
|
||||||
|
|
||||||
Returns:
|
Returns:
|
||||||
True if package is installed and version >= min_version, False otherwise
|
True if package is installed and version >= min_version, False otherwise
|
||||||
|
|||||||
@@ -5,7 +5,7 @@ set -euxo pipefail
|
|||||||
# Set up environment variables
|
# Set up environment variables
|
||||||
IS_BLACKWELL=${IS_BLACKWELL:-0}
|
IS_BLACKWELL=${IS_BLACKWELL:-0}
|
||||||
CU_VERSION="cu129"
|
CU_VERSION="cu129"
|
||||||
FLASHINFER_VERSION=0.6.2
|
FLASHINFER_VERSION=0.6.3
|
||||||
OPTIONAL_DEPS="${1:-}"
|
OPTIONAL_DEPS="${1:-}"
|
||||||
|
|
||||||
# Detect system architecture
|
# Detect system architecture
|
||||||
|
|||||||
Reference in New Issue
Block a user