Skip to content

[skyrl-train] Upgrade vllm to 0.13.0 (and torch to 2.9.0) for both vllm and mcore extras #887

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 2 commits into
base: main
Choose a base branch
from
+461 −318
Open

[skyrl-train] Upgrade vllm to 0.13.0 (and torch to 2.9.0) for both vllm and mcore extras #887

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
56ae05b
x
erictang000 Jan 15, 2026
52015fa
x
erictang000 Jan 15, 2026
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
12 changes: 6 additions & 6 deletions skyrl-train/pyproject.toml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -125,9 +125,9 @@ sandboxes = [
"litellm[proxy]>=1.67.5",
]
vllm = [
"vllm==0.11.0",
"vllm==0.13.0",
"flash-attn==2.8.3",
"torch==2.8.0",
"torch==2.9.0",
"flashinfer-python",
"flashinfer-jit-cache",
"torchvision"
Expand All @@ -142,13 +142,13 @@ sglang = [
mcore = [
"transformer-engine[pytorch]==2.9.0",
"flash-attn==2.8.1",
"vllm==0.11.0",
"torch==2.8.0",
"flashinfer-python==0.5.2",
"vllm==0.13.0",
"torch==2.9.0",
"flashinfer-python==0.5.3",
"torchvision",
"megatron-bridge @ git+https://github.com/NVIDIA-NeMo/Megatron-Bridge.git@v0.2.0",
"megatron-core==0.15.0",
"flashinfer-jit-cache==0.5.2",
"flashinfer-jit-cache==0.5.3",
"nvidia-modelopt",
]
flashrl = [
Expand Down
15 changes: 12 additions & 3 deletions skyrl-train/skyrl_train/inference_engines/vllm/vllm_engine.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -364,27 +364,36 @@ def _create_engine(self, *args, **kwargs):
model_name = model_path

base_model_paths = [BaseModelPath(name=model_name, model_path=model_path)]
models = OpenAIServingModels(engine, model_config, base_model_paths)

# vllm >= 0.11.2 removed model_config from OpenAI serving APIs
is_new_api = version.parse(vllm.__version__) >= version.parse("0.11.2")
legacy_kwargs = {}
if is_new_api:
models = OpenAIServingModels(engine, base_model_paths)
else:
models = OpenAIServingModels(engine, model_config, base_model_paths)
legacy_kwargs["model_config"] = model_config

# TODO(Charlie): revisit kwargs `enable_auto_tools` and `tool_parser` when we need to
# support OAI-style tool calling; and `request_logger` for better debugging.
self.openai_serving_chat = OpenAIServingChat(
engine_client=engine,
model_config=model_config,
models=models,
response_role="assistant",
request_logger=None,
chat_template=None,
chat_template_content_format="auto",
**legacy_kwargs,
**openai_kwargs,
)

# TODO(Charlie): revisit kwargs `return_tokens_as_token_ids`,
# `enable_prompt_tokens_details`, `enable_force_include_usage`.
self.openai_serving_completion = OpenAIServingCompletion(
engine_client=engine,
model_config=model_config,
models=models,
request_logger=None,
**legacy_kwargs,
)
return engine

Expand Down
Loading