Skip to content

prime-runs #3459

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
kcoopermiller wants to merge 6 commits into
base: main
Choose a base branch
from
Draft

prime-runs #3459

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
9c561de
feat(monitors): log to the platform through prime-runs
kcoopermiller Sep 1, 2026
8d7837e
chore: update prime-runs
kcoopermiller Sep 1, 2026
ba3e2fe
Merge branch 'main' into cooper/prime-runs
kcoopermiller Sep 1, 2026
8b443c5
Merge branch 'main' into cooper/prime-runs
kcoopermiller Sep 2, 2026
d1eeeea
monitors: pin the current prime-runs, adopt the platform run id, add …
kcoopermiller Sep 2, 2026
33f842c
docs: keep the platform-monitoring intro line as it was
kcoopermiller Sep 2, 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
4 changes: 2 additions & 2 deletions docs/training.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -384,9 +384,9 @@ Or set it in TOML:
name = "my-experiment"
```

Every 10th step the orchestrator uploads the step's episodes (full conversations with rewards and advantages) to the run's sample viewer.
The monitor is a thin layer over the [`prime-runs`](https://github.com/PrimeIntellect-ai/prime/tree/main/packages/prime-runs) SDK (installed as `prime-runs[train]`): it registers the run, streams per-step metrics, uploads every 10th step's episodes (full conversations with rewards and advantages) to the run's sample viewer, and closes the run out. A process that exits without finishing is reported as crashed. Episodes are stamped with the platform run's id, so the run's traces can be queried by the id the dashboard shows; W&B keeps the launcher's `PRL_RUN_ID`.

Requires `PRIME_API_KEY` (set via `prime login` or env var) and an allowlisted team. Currently internal-only.
Requires `PRIME_API_KEY` (`prime login` or the env var) and a team (`PRIME_TEAM_ID`, or the team selected with `prime login`) enabled for external runs. A configured monitor must work: a missing key or a team outside the allowlist fails the launch. `PRIME_RUNS_MODE=disabled` keeps the monitor configured but opens no platform run; `RUN_ID=<id>` attaches to an external run a launcher already created instead of registering a new one. Currently internal-only.

## Rules of Thumb

Expand Down
5 changes: 5 additions & 0 deletions pyproject.toml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -17,6 +17,7 @@ dependencies = [
"wandb>=0.26.1",
"wandb-workspaces>=0.4.3",
"prime>=0.6.19",
"prime-runs[train]>=0.1.0",
"pyzmq>=27.1.0",
"aiolimiter>=1.2.1",
"tenacity>=8.2.0",
Expand Down Expand Up @@ -221,6 +222,8 @@ flash_attn_3 = false
# appear in `uv tree` need to be listed; the resolver ignores entries for
# packages it never sees.
prime = false
prime-runs = false
prime-traces = false
prime-sandboxes = false
prime-tunnel = false
prime-evals = false
Expand All @@ -237,6 +240,8 @@ torchao = false
prime-kernels = false

[tool.uv.sources]
# TEMPORARY: prime-runs is not on PyPI yet (PrimeIntellect-ai/prime #856); drop once 0.1.0 is published
prime-runs = { git = "https://github.com/PrimeIntellect-ai/prime.git", subdirectory = "packages/prime-runs", rev = "aba26790" }
prime-rl-configs = { path = "packages/prime-rl-configs", editable = true }
# prime-rl consumes `verifiers.v1` (incl. the built-in harnesses/tasksets); env tasksets
# live under deps/{verifiers,prime-envs}/environments.
Expand Down
Loading