Skip to content

Fix torch compile cache dir conflicts when enforce_eager=False #368

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

Merged
chenyushuo merged 2 commits into from
Nov 6, 2025
Merged

Fix torch compile cache dir conflicts when enforce_eager=False #368

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
7b56ea7
fix conflicts when multiple vllm instances are started simultaneously
pan-x-c Nov 5, 2025
17cb598
update enforce eager defaut to false
pan-x-c Nov 6, 2025
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
2 changes: 1 addition & 1 deletion benchmark/config/countdown-template.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -54,7 +54,7 @@ explorer:
rollout_model:
engine_num: 2
tensor_parallel_size: 1
enforce_eager: true
enforce_eager: false
enable_prefix_caching: false
enable_chunked_prefill: false
gpu_memory_utilization: 0.9
Expand Down
2 changes: 1 addition & 1 deletion docs/sphinx_doc/source/tutorial/example_step_wise.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -140,7 +140,7 @@ explorer:
engine_num: 2
tensor_parallel_size: 2
enable_prefix_caching: false
enforce_eager: true
enforce_eager: false
dtype: bfloat16
seed: 42
gpu_memory_utilization: 0.7
Expand Down
2 changes: 1 addition & 1 deletion docs/sphinx_doc/source_zh/tutorial/example_step_wise.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -135,7 +135,7 @@ explorer:
engine_num: 2
tensor_parallel_size: 2
enable_prefix_caching: false
enforce_eager: true
enforce_eager: false
dtype: bfloat16
seed: 42
gpu_memory_utilization: 0.7
Expand Down
2 changes: 1 addition & 1 deletion examples/agentscope_react/gsm8k.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -43,7 +43,7 @@ explorer:
engine_num: 4
tensor_parallel_size: 1
enable_prefix_caching: false
enforce_eager: true
enforce_eager: false
enable_openai_api: true
enable_history: true
enable_auto_tool_choice: true
Expand Down
2 changes: 1 addition & 1 deletion examples/agentscope_tool_react/agentscopev0_tool_react_dapo.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -44,7 +44,7 @@ explorer:
engine_num: 4
tensor_parallel_size: 1
enable_prefix_caching: false
enforce_eager: true
enforce_eager: false
enable_openai_api: true
enable_history: true
dtype: bfloat16
Expand Down
2 changes: 1 addition & 1 deletion examples/agentscope_tool_react/agentscopev0_tool_react_gsm8k.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -44,7 +44,7 @@ explorer:
engine_num: 4
tensor_parallel_size: 1
enable_prefix_caching: false
enforce_eager: true
enforce_eager: false
enable_openai_api: true
enable_history: true
dtype: bfloat16
Expand Down
2 changes: 1 addition & 1 deletion examples/agentscope_tool_react/agentscopev1_tool_react_dapo.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -44,7 +44,7 @@ explorer:
engine_num: 4
tensor_parallel_size: 1
enable_prefix_caching: false
enforce_eager: true
enforce_eager: false
enable_openai_api: true
enable_history: true
dtype: bfloat16
Expand Down
2 changes: 1 addition & 1 deletion examples/agentscope_websearch/agentscopev1_websearch_agent.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -68,7 +68,7 @@ explorer:
engine_num: 4
tensor_parallel_size: 1
enable_prefix_caching: false
enforce_eager: true
enforce_eager: false
dtype: bfloat16
seed: 42
gpu_memory_utilization: 0.7
Expand Down
2 changes: 1 addition & 1 deletion trinity/common/config.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -447,7 +447,7 @@ class InferenceModelConfig:
engine_num: int = 1
tensor_parallel_size: int = 1
use_v1: bool = True
enforce_eager: bool = True
enforce_eager: bool = False
enable_prefix_caching: bool = False
enable_chunked_prefill: bool = False
gpu_memory_utilization: float = 0.9
Expand Down
7 changes: 7 additions & 0 deletions trinity/common/models/vllm_model.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -53,6 +53,13 @@ def __init__(
os.environ["VLLM_ENABLE_V1_MULTIPROCESSING"] = "0"
if get_vllm_version() >= parse_version("0.11.0"):
os.environ["VLLM_ALLREDUCE_USE_SYMM_MEM"] = "0"
if not config.enforce_eager:
# To avoid torch compile conflicts when multiple model are started simultaneously.
# remove this when the following PR is released:
# https://github.com/vllm-project/vllm/pull/27616
os.environ["VLLM_CACHE_ROOT"] = os.path.expanduser(
f"~/.cache/vllm/{config.bundle_indices}"
)
self.default_sampling_params = vllm.SamplingParams(
n=1,
temperature=0.0,
Expand Down