Run the first 100 steps of Run 0 on one GPU and report results
Maintainers usually reply within 1 day
Nobody has claimed this yet.
Assessment
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Newbie friendliness
- 42/100
- Issue type
- Feature
- Clarity
- Clearly specified
- Activity status
- Active
- Domain
- distributed-systems, machine-learning
Research direction
Start with docs/run0.md and inspect the launcher in src/ofl/run0.py; follow the documented setup and run uv run ofl run0 launch --dry-run before using a one-GPU node. Record the hardware, make doctor output, exact command, throughput, peak memory, and losses at steps 1, 50, and 100. Done when runs/run0-first-100/README.md reports results or failures and make check and make test pass.
Written by the indexing model from the issue text.
Description
Context
Run 0 (RFC 0001 Rung 0) has a launcher, config and pinned upstream, but has never run on a GPU (README status).
Task
On one NVIDIA GPU node, follow docs/run0.md, run the first 100 steps of the 1B config (reduce batch size if memory requires and record it), and report: hardware, exact command, tokens/s, loss curve, peak memory, and anything that broke. Use the data the doc specifies or the smallest shard that works and say which.
Definition of done
- A PR adding
runs/run0-first-100/README.mdwith the command, environment (make doctoroutput), log excerpt and loss at steps 1, 50, 100 - Any doc or launcher fixes the run needed, in the same PR or linked
- Failures reported honestly; a failed run is a valid result
-
make checkandmake testpass - Commits signed off (
git commit -s)
Pointers
docs/run0.md,src/ofl/run0.py,uv run ofl run0 launch --dry-run- OLMo-core
- Ask in Discord
#training, or comment here. Comment "I'll take this" to be assigned. A draft PR after an hour of work is welcome.
Skills: ML, a GPU with 24 GB+
Agent: omp / claude-opus-5-5 (issue drafted by an agent; a maintainer reviews the PR)
- Dominant language
- Python
- Stars
- 1
- Forks
- 2
- Avg merge
- 2h 4m
- Merged PRs (30d)
- 23
Getting set up
- No Dockerfile or Docker Compose file
- No pull request template
- Read the contributing guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
More from OpenFrontierLabs/openfrontier
-
agent-authored area/evals good first issue help wanted layer/8-outputs type/feature
Difficulty 2/5 1-3 hours Newbie friendliness 72/100
OpenFrontierLabs/openfrontier#26 ·
Maintainers usually reply within 1 day
-
agent-authored area/docs good first issue help wanted layer/7-hosting type/docs
Difficulty 3/5 1-2 days Newbie friendliness 77/100
OpenFrontierLabs/openfrontier#25 ·
Maintainers usually reply within 1 day
-
agent-authored area/web good first issue help wanted layer/6-console type/feature
Difficulty 3/5 1-2 days Newbie friendliness 68/100
OpenFrontierLabs/openfrontier#24 ·
Maintainers usually reply within 1 day
-
Add hash and licence fields for every entry in `manifests/datasets.toml`, checked by `ofl check`Possibly taken @Neeraj111010 claimed this 1 day ago. Openagent-authored area/data area/python good first issue help wanted layer/5-storage type/feature
Difficulty 3/5 1-2 days Newbie friendliness 76/100
OpenFrontierLabs/openfrontier#23 · 1 comment ·
Maintainers usually reply within 1 day
-
agent-authored area/data good first issue help wanted layer/5-storage type/feature
Difficulty 4/5 3-5 days Newbie friendliness 62/100
OpenFrontierLabs/openfrontier#22 ·
Maintainers usually reply within 1 day
All issues in OpenFrontierLabs/openfrontier
Similar issues
-
Email validation accepts a trailing newlinePossibly taken @marcello-pichini claimed this today. Open
Difficulty 2/5 1-3 hours Newbie friendliness 78/100
Maintainers usually reply within 1 day
-
Difficulty 2/5 1-3 hours Newbie friendliness 76/100
Maintainers usually reply within 1 day
-
Difficulty 2/5 1-3 hours Newbie friendliness 72/100
dbt-labs/spark-utils#51 ·
-
Difficulty 2/5 1-3 hours Newbie friendliness 85/100
Maintainers usually reply within 1 day
-
bug frontend
Difficulty 2/5 1-3 hours Newbie friendliness 88/100
vendrabuck/HERD#1419 ·
Maintainers usually reply within 1 day