Skip to content

Add a transformers-agentic example corpus#42

Merged
dacorvo merged 2 commits into
mainfrom
feat/transformers-agentic
Jun 26, 2026
Merged

Add a transformers-agentic example corpus#42
dacorvo merged 2 commits into
mainfrom
feat/transformers-agentic

Conversation

@dacorvo

@dacorvo dacorvo commented Jun 26, 2026

Copy link
Copy Markdown
Collaborator

No description provided.

dacorvo and others added 2 commits June 26, 2026 14:24
`run --label <s>` records a free-form label in run.json; export emits it as a
constant `label` column (read from run.json the same way as `agent`). Lets a
comparison sweep tag each run with its condition (e.g. a bare/clone/skill tier)
so cells stay identifiable in `agentcap ls`, inspect, and the published dataset.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Port of huggingface/is-it-agentic-enough: 16 prompts that each ask an agent to
run a named Hugging Face model and report the result (sentiment, NER, QA,
transcription, captioning, …), across the bare/clone/skill assistance tiers.

build-toolenv.sh builds a self-contained, relocatable transformers bundle
inside ubuntu:24.04 (the agent-image base) — pinned to the unreleased
agentic-CLI commit, CPU torch — and prewarms every corpus model into a shared
cache read offline at run time. run.sh mounts it via --tool-dir and seeds the
per-tier sandbox. Used to compare models/agents through agentcap's capture
path; match/marker scoring stays with the upstream harness.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
@dacorvo dacorvo merged commit e917428 into main Jun 26, 2026
7 checks passed
@dacorvo dacorvo deleted the feat/transformers-agentic branch June 26, 2026 16:54
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant