Skip to content

[chore] bump vllm to 0.24.0 #1854

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
erictang000 wants to merge 1 commit into
base: main
Choose a base branch
from
Open

[chore] bump vllm to 0.24.0 #1854

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
10 changes: 5 additions & 5 deletions pyproject.toml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -105,7 +105,7 @@ skyrl-train = [

fsdp = [
"skyrl[skyrl-train]",
"vllm==0.23.0; sys_platform == 'linux'",
"vllm==0.24.0; sys_platform == 'linux'",
"vllm-router; sys_platform == 'linux'",
# The `nixl` shim provides that namespace and dispatches to `nixl_cu12`.
# `nixl-cu12` ships the `nixl_cu12` module, but vLLM imports `nixl._api`.
Expand All @@ -129,7 +129,7 @@ megatron = [
"flash-linear-attention; sys_platform == 'linux'",
"causal-conv1d; sys_platform == 'linux'",
"mamba-ssm>=2.3.0; sys_platform == 'linux'",
"vllm==0.23.0; sys_platform == 'linux'",
"vllm==0.24.0; sys_platform == 'linux'",
"vllm-router; sys_platform == 'linux'",
# The `nixl` shim provides that namespace and dispatches to `nixl_cu12`.
# `nixl-cu12` ships the `nixl_cu12` module, but vLLM imports `nixl._api`.
Expand Down Expand Up @@ -236,7 +236,7 @@ override-dependencies = [
# so it doesn't pull the CUDA-13 stack and bump torch off the cu12 pin.
"nixl-cu13; sys_platform == 'never'",
# Megatron-Bridge pins flashinfer-python==0.6.8.post1, which conflicts with
# our exact 0.6.12 pin (the version vLLM 0.23.0 requires). Override it to our version.
# our exact 0.6.12 pin (the version vLLM 0.24.0 requires). Override it to our version.
"flashinfer-python==0.6.12; sys_platform == 'linux' and platform_machine == 'x86_64'",
"flashinfer-jit-cache==0.6.12; sys_platform == 'linux' and platform_machine == 'x86_64'",
"flashinfer-cubin==0.6.12; sys_platform == 'linux' and platform_machine == 'x86_64'",
Expand Down Expand Up @@ -281,14 +281,14 @@ explicit = true

[[tool.uv.index]]
name = "vllm-cu129"
url = "https://wheels.vllm.ai/0.23.0/cu129"
url = "https://wheels.vllm.ai/0.24.0/cu129"
explicit = true

[tool.uv.sources]
skyrl-gym = { path = "./skyrl-gym", editable = true }
# Match torch's CUDA variant (cu128).
flashinfer-jit-cache = { index = "flashinfer-cu128", marker = "sys_platform == 'linux'" }
# vllm 0.23.0's PyPI wheel needs CUDA 13 (libcudart.so.13); the cu129 wheel
# vllm 0.24.0's PyPI wheel needs CUDA 13 (libcudart.so.13); the cu129 wheel
# links libcudart.so.12, which torch+cu128 supplies.
vllm = [
{ index = "vllm-cu129", marker = "sys_platform == 'linux'" },
Expand Down
Loading
Loading