GREP (GPU)

ACTIVE
computer-science

GREP — The Great Research Extraction Project. Extracts software mentions from scientific papers using a five-way DeBERTa ensemble (the Extract2 model). Each work unit is one PDF. Runs unbounded — no wall-clock deadline; liveness is heartbeat-based. GPU build (fp16, validated baseline); requires an NVIDIA GPU with ~22 GB VRAM.

How to Contribute
  1. Install the Lettuce volunteer CLI
  2. Run: lettuce-volunteer init
  3. Run: lettuce-volunteer attach --server infra.scios.tech
  4. Run: lettuce-volunteer start

Your contributions will be automatically tracked and credited. Learn more about volunteering

Created by
@admin

Administrator

Member since May 2026

Runtime
Container
ghcr.io/jring-o/extract2-lettuce:1.3-gpu
GPU Required (NVIDIA) · 18432 MB VRAM
Resource Requirements
4 CPU cores
33.2 GB RAM
122880 MB disk
Statistics
Active Volunteers0
Work Units Completed0 / 0
Total Credit0
Avg. Completion Time—
Task Pattern
Custom

Uses researcher-uploaded work units with custom input data.

Custom work unit upload via API