OpenThoughts
Official@open-thoughts
Open collaborations on data-centric research
Agent Skills by OpenThoughts
Showing 27 vetted skills indexed across 1 GitHub repositories.
open-thoughtsdatagen-launch-iris
Orchestrate datagen jobs on Marin's Iris TPU cluster from launch to cleanup.
open-thoughtsrl-agentic-launch-jupiter
Launch agentic RL experiments on Jupiter with SkyRL, Harbor, and Daytona.
open-thoughtsmonitor-job-tables
Format HPC job-status data into bucketed box-drawing tables.
open-thoughtsmonitor-restore
Restore the 3-hour Jupiter+Leonardo cluster sweep loop via /loop 3h or CronCreate.
open-thoughtsmonitor-restore-iris-cron
Re-register the every-3-hours Iris job-monitor cron via CronCreate.
open-thoughtsdatagen-job-cleanup
Upload verified datagen traces to HF and free disk space.
open-thoughtssft-job-cleanup
Publish finished LLaMA-Factory SFT jobs to Hugging Face and clean up HPC cluster state.
open-thoughtseval-agentic-launch-iris
Manages the full lifecycle of Iris evaluation jobs via the command line.
open-thoughtscrud-otagent-supabase
Query, aggregate, and safely write OT-Agent eval data in Supabase.
open-thoughtssft-cleanup-hf-only
Publish HF-only SFT cells to Hugging Face without database registration.
open-thoughtseval-standard-cleanup
Consolidate finished standard eval jobs into a SCORES.md tracker.
open-thoughtssft-launch-jupiter
Launch SFT experiments on Jupiter via hpc.launch with LLaMA-Factory.
open-thoughtseval-agentic-cleanup
Audit and remediate agentic eval runs across SLURM workflows.
open-thoughtsdatagen-launch
Launch datagen trace-generation jobs on HPC clusters and upload trajectories to HuggingFace.
open-thoughtsrl-job-cleanup
Publish the best RL training checkpoint to Hugging Face and Supabase after job termination.
open-thoughtscode-execute-staged-plan
Execute staged codebase plans with per-stage validation gates and agent_logs traceability.
open-thoughtsdatagen-reduce-dataset-snapshots
Reduce Daytona snapshot counts by grouping Dockerfiles into shared environments.
open-thoughtsanalyze-rl-behavior
Synthesize evaluation traces, training logs, and GPT-5 assessments into a structured behavioral report.
open-thoughtssupervisor-init
Bootstrap the lab supervisor role to orchestrate subagents, crons, and tools.
open-thoughtsrl-standard-launch-leonardo
Launch and manage offline SkyRL experiments on CINECA Leonardo via sbatch wrappers.
open-thoughtsmonitor-cron-sweep
Consolidate squeue and sacct job statuses into a unified cross-cluster report.
open-thoughtsdatagen-standard-launch
Generate synthetic datasets via Curator sharding or declarative data scripts.
open-thoughtscode-create-staged-plan
Generate dependency-ordered staged plans for large codebase changes.
open-thoughtseval-standard-launch
Automate Delphi 6279 RL-scaling-laws evaluation workflows on Leonardo with sbatch submission.
Frequently Asked Questions About OpenThoughts
FAQPage SchemaWhat specific tasks can I perform using OpenThoughts?▼
You can orchestrate synthetic data generation, launch LLaMA-Factory supervised fine-tuning jobs, execute reinforcement learning experiments, and manage cluster-wide evaluation lifecycles. It provides specific capabilities for job cleanup, Hugging Face publication, and cross-cluster status monitoring using SLURM and Cron-based scheduling.
Which technical personas benefit from these capabilities?▼
These capabilities are designed for research engineers and machine learning practitioners managing large-scale training runs on HPC clusters. It is ideal for those requiring structured oversight of distributed compute resources, dataset versioning, and automated evaluation reporting for agentic models.
What are the prerequisites for running these jobs?▼
Execution requires access to HPC environments like CINECA Leonardo or Jupiter, configured with SLURM for job submission. Dependencies include LLaMA-Factory for fine-tuning, Supabase for metadata storage, and authenticated access to Hugging Face for publishing verified training traces and model checkpoints.