bc815b3272
* feat(byo): apply-swap — download + serve a Route-C fine-tune
A curated-arch fine-tune (e.g. josefprusa/ThinkingCap-Qwen3.6-27B AutoRound
→ qwen3.6-27b) hard-stops at the pull gate's stratum-5 no-fit-model (correct
— nothing to price), so the c3 Bring [D] download failed with "download did
not complete". Route-C was pure guidance ("clone the compose, point --model
at your weights") with no code behind it — the deferred "swap_path apply".
Build it as a DISTINCT action that never touches the locked 6-stratum gate:
- scripts/lib/profiles/swap_apply.py (new): resolve the Route-C swap
(arch→curated sibling + has_mtp_head from deriver), download the brought
weights SHA-verified via downloader.download_model (a minimal EInput), and
emit a serve-locally compose that CLONES the --profile-like sibling's REAL
compose (keeping its curated chat-template / reasoning+tool parsers / MTP
wiring — NOT the derived-vllm template, which drops all that) with --model
re-pointed at a `<weights>:/brought-model:ro` mount, --served-model-name set
to the brought basename, and --speculative-config kept iff the checkpoint
carries an MTP head (else dropped). Written alongside the sibling compose as
`_brought-<san>.yml` (gitignored) so its relative ../ mounts resolve.
- scripts/pull.sh: intercept `--apply-swap` in the wrapper (strictly additive,
like --json) → run the apply-swap action; prints `[apply-swap] compose: …`.
The gate path is byte-unchanged without the flag.
- c3: services.run_bring_download gains `apply_swap=` (appends --apply-swap +
captures the emitted compose via last_swap_compose()); the [D] worker passes
it on route==C and stashes the compose; ② Serve serves that swap compose
directly (serve_generated) instead of a reproduction of the sibling's own
catalog compose.
- pull.py: fix the stale NOTE — "the bf16 base won't fit and lacks the MTP
head" → the base HAS the head; now just the size reason.
Tests: test-pull-swap.sh section 4 (emit: --model repointed, MTP kept/dropped
by head presence, curated flags + JSON args survive the YAML round-trip,
distinct container_name); test_services apply_swap flag + compose capture. The
locked-gate hard-stop tests stay green; --json fit-check byte-identical.
Validated live: pull.sh --json still no-fit-model+route-C, and apply_swap emits
a correct ThinkingCap swap compose (MTP kept).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01EfF565T9eSLaqGzidyJ1Pm
* c3: reframe Route-C fit-check card as ✓ Servable + a clear [D] action
The fit-check card for a curated-arch fine-tune (Route-C swap) was
self-contradicting: a RED "not eligible" + a bare "no-fit-model" token at
the top, then a GREEN "② Serve is armed with <sibling>" at the bottom —
which read as a dead-end AND named the wrong model (the sibling, not the
brought fine-tune). The engine's "no-fit-model" is the *means* (generic
fit-math can't price a curated-hybrid arch), not the user's answer; the
outcome is servable via the sibling's recipe.
Reframe the Route-C case in _byo_result_text:
- green "✓ Servable — a fine-tune of <sibling>" instead of red "not eligible",
- one plain "How it serves" line (reuses the sibling's chat-template / tools /
spec-dec with your weights) instead of the --quantization/impl bullets,
- MTP kept/dropped line from has_mtp_head,
- the next-step is an explicit "→ Press [D] to download + serve <BROUGHT model>"
(the fine-tune, not the sibling) — which #628's --apply-swap now fulfils,
- raw verdict + backend note dimmed for debugging, not the headline.
Non-swap cases (eligible / Route A / B / plain no-fit) are byte-unchanged.
Regression test added; c3 fast suite green.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01EfF565T9eSLaqGzidyJ1Pm
---------
Co-authored-by: noonghunna <10742901+noonghunna@users.noreply.github.com>
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
73 lines
2.3 KiB
Plaintext
73 lines
2.3 KiB
Plaintext
# Model weights — never commit. Default download dir is models-cache/ (set
|
|
# via MODEL_DIR env var if you want a different path).
|
|
models-cache/
|
|
*.safetensors
|
|
*.gguf
|
|
*.bin
|
|
|
|
# Genesis patches fetched by setup.sh (upstream, not ours to vendor).
|
|
# Lives at models/<model>/vllm/patches/genesis/ in the new layout.
|
|
# No trailing slash — also matches local symlinks pointing at out-of-tree clones.
|
|
models/*/vllm/patches/genesis
|
|
|
|
# v0.8.0 [E] Pull-Emit-Derived runtime capture artifacts (the §6 capture
|
|
# points + manifest; consumed by the later [F] Loop phase). Per-run output,
|
|
# never committed.
|
|
.pull-captures/
|
|
|
|
# serve-cockpit producer lane ② Serve — temp composes generated by
|
|
# generate-compose.sh (CockpitData.generate_compose). Written under the repo
|
|
# root so their relative ../ mounts resolve; cleaned up on serve/decline, but
|
|
# gitignored so a crash-leftover never lands in `git add -A`.
|
|
c3-genc-*.yml
|
|
|
|
# BYO Route-C apply-swap serve-locally composes (pull.sh --apply-swap): a clone
|
|
# of the sibling compose with --model → the brought weights, written IN the
|
|
# sibling's compose dir so its relative ../ mounts resolve. Generated artifact.
|
|
_brought-*.yml
|
|
|
|
# Python
|
|
__pycache__/
|
|
*.pyc
|
|
*.pyo
|
|
*.egg-info/
|
|
.venv/
|
|
venv/
|
|
|
|
# Editor
|
|
.vscode/
|
|
.idea/
|
|
*.swp
|
|
.DS_Store
|
|
|
|
# Docker compose state
|
|
**/compose/*.log
|
|
|
|
# Local env overrides
|
|
.env
|
|
*.env.local
|
|
|
|
# Local safety backups of compose files (e.g. docker-compose.yml.default)
|
|
**/compose/*.default
|
|
|
|
# Patched vLLM source clone (volume-mounted by dual-card composes)
|
|
vllm-src/
|
|
|
|
# Internal diagnostics / feasibility memos — kept local, not published.
|
|
# Per-model published diagnostics live at models/<name>/vllm/diagnostics/.
|
|
docs/diagnostics/
|
|
|
|
# Bench output. Most runs are intermediate investigations (cliff probes, soak
|
|
# validation matrices, residency instrumentation) — those live in issue threads
|
|
# and gitignored docs/diagnostics/. Bench evidence backing numbers in BENCHMARKS.md
|
|
# / STRUCTURED_COT.md / CHANGELOG is tracked via the negations below; future
|
|
# grammar-ab-* and grammar-full-* runs will auto-track when committed.
|
|
results/*
|
|
!results/grammar-ab-*/
|
|
!results/grammar-full-*/
|
|
!results/lucebox-*/
|
|
!results/v0.20-migration/
|
|
# #252: curated quality-baseline corpus (committed; the trusted n>=3 aggregates
|
|
# `quality-baseline.sh` diffs fresh runs against). Run output elsewhere stays ignored.
|
|
!results/baselines/
|