[FlashInfer] Bump FlashInfer version from 0.6.2 to 0.6.3 (#18448)
This commit is contained in:
committed by
GitHub
parent
710d873ba6
commit
1be41e9036
@@ -28,8 +28,8 @@ dependencies = [
|
||||
"datasets",
|
||||
"einops",
|
||||
"fastapi",
|
||||
"flashinfer_python==0.6.2", # keep it aligned with jit-cache version in Dockerfile
|
||||
"flashinfer_cubin==0.6.2",
|
||||
"flashinfer_python==0.6.3", # keep it aligned with jit-cache version in Dockerfile
|
||||
"flashinfer_cubin==0.6.3",
|
||||
"gguf",
|
||||
"hf_transfer",
|
||||
"huggingface_hub",
|
||||
|
||||
@@ -841,7 +841,7 @@ def _set_envs_and_config(server_args: ServerArgs):
|
||||
if server_args.attention_backend == "flashinfer":
|
||||
assert_pkg_version(
|
||||
"flashinfer_python",
|
||||
"0.6.2",
|
||||
"0.6.3",
|
||||
"Please uninstall the old version and "
|
||||
"reinstall the latest version by following the instructions "
|
||||
"at https://docs.flashinfer.ai/installation.html.",
|
||||
|
||||
@@ -1616,8 +1616,8 @@ class ServerArgs:
|
||||
and self.moe_a2a_backend == "none"
|
||||
and self.moe_runner_backend == "auto"
|
||||
):
|
||||
# Only enable flashinfer_trtllm if flashinfer-python version is >= 0.6.2
|
||||
if check_pkg_version_at_least("flashinfer-python", "0.6.2"):
|
||||
# Only enable flashinfer_trtllm if flashinfer-python version is >= 0.6.3
|
||||
if check_pkg_version_at_least("flashinfer-python", "0.6.3"):
|
||||
self.moe_runner_backend = "flashinfer_trtllm"
|
||||
logger.info(
|
||||
"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:
|
||||
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:
|
||||
True if package is installed and version >= min_version, False otherwise
|
||||
|
||||
Reference in New Issue
Block a user