Skip to content

feat(v1): add Trace.reward_shaping for trainer reward-shaping terms #2436

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

Draft
hallerite wants to merge 1 commit into
base: main
Choose a base branch
from
Draft

feat(v1): add Trace.reward_shaping for trainer reward-shaping terms #2436

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
4 changes: 4 additions & 0 deletions verifiers/v1/trace.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -404,6 +404,10 @@ class Trace(BaseModel, Generic[DataT, StateT, AgentConfigT]):
preceding error)."""
metrics: dict[str, float | None] = Field(default_factory=dict)
"""Unweighted, named metrics; `None` as in `rewards`."""
reward_shaping: dict[str, float] = Field(default_factory=dict)
"""Additive training-reward terms recorded by a consumer's reward shapers (a length
penalty, ...), keyed by shaper name. A training annotation like `MessageNode.advantages`:
`reward` stays the env's verdict; the trainer scores against `reward` plus these terms."""
info: dict[str, Any] = Field(default_factory=dict)
"""Scratch space for task-specific metadata."""
state: StateT = Field(default_factory=State, exclude=True)
Expand Down