zytsyj/dsh-gpu

README 摘要

dsh-gpu GPU-aware execution layer for DeepSeek Harness (dsh). Out-of-tree plugin; no harness patches required. Agents get three tools — gpu status , gpu exec , gpu run bg — plus an optional per-step GPU context line. Cards are selected automatically (freest first) with CUDA VISIBLE DEVICES set in the command environment; pin a card explicitly when you care. How it works - gpu status — one query, every device: memory used/total, SM utilization, temperature, and a free/busy verdict. A device is busy at or above 80% memory used or 50% utilization (both configurable). - gpu exec — one-shot command with a selected card: CUDA VISIBLE DEVICES= is passed through the mounted ctx.shell executor's environment. Auto-select or pin gpuIndex ; select count cards for multi-GPU commands. - gpu run bg — long-running GPU jobs (training, inference servers, benchmarks) register as a gpu job in ctx.jobs : returns a job id immediately, read with job output , stop with job kill . - Per-step context (optional, on by default) — injects a one-line GPU snapshot into eligible steps (the time-context pattern), rate-limited to one sample per minute. All execution rides the mounted shell executor . Local host, or…

分类