Lap Tran
81ccf548e5
CI / build (push) Has been cancelled
Batch 1 (High severity): - Bug 1: --audit cat3 now checks .automaton/tasks/ paths - Bug 4: Verdict PASS/FAIL uses structured ## Status: line parsing - Bug 5: register-guards.sh checks .json/.jsonc, writes plugin key, strips comments - Bug 7: --can-edit/--scope-check path prefix uses os.sep boundary Batch 2 (Medium/Low severity): - Bug 2: migrate-project.sh find command parentheses for -prune binding - Bug 3: vram_detect model prefix matching with known-suffix whitelist - Bug 6: dashboard reads .state file before artifact heuristic fallback - Bug 8: removed wildcard CORS, added security headers (nosniff, DENY) - Bug 9: stale-task detection uses .state.lastedit instead of .state mtime - Bug 10: TEST_PLAN.md maps to test_design (was implement) 249 tests pass (up from 235). All 10 tasks driven through full workflow to completion.
25 lines
1.4 KiB
Markdown
25 lines
1.4 KiB
Markdown
# Implementation: fix-vram-model-prefix-match
|
|
|
|
## Bug
|
|
`_lookup_model_context()` in `vram_detect.py` used raw `startswith()` for model name matching, causing false positives like `phi-4` matching `phi-40` or `phi-4-mini-instruct` (a different model with different context window).
|
|
|
|
## Fix
|
|
Replaced the raw `startswith()` with a three-tier matching strategy:
|
|
1. **Exact match** — `name_lower == key_lower`
|
|
2. **Ollama parameter tag** — `name_lower.startswith(key_lower + ":")` (e.g. `deepseek-r1:7b` matches `deepseek-r1`)
|
|
3. **Known instruction-tuning suffix** — `name_lower.startswith(key_lower + "-")` only if the next segment is in `_KNOWN_MODEL_SUFFIXES = {"instruct", "chat", "it", "fp16", "f16", "bf16"}` (e.g. `llama-3.1-8b-instruct` matches `llama-3.1-8b`)
|
|
|
|
Keys are sorted by length descending so the most specific match wins first.
|
|
|
|
This prevents false matches:
|
|
- `phi-4-mini-instruct` → `mini` not in known suffixes → no match ✓
|
|
- `gpt-4o-foo-unknown` → `foo` not in known suffixes → no match ✓
|
|
- `phi-40` → no `:` or known-suffix separator → no match ✓
|
|
|
|
## Files Changed
|
|
- `scripts/vram_detect.py`: Added `_KNOWN_MODEL_SUFFIXES` set, rewrote `_lookup_model_context()` with three-tier matching
|
|
|
|
## Tests
|
|
- `test_lookup_model_context_no_false_prefix_match`: Asserts `phi-4-mini-instruct` and `gpt-4o-foo-unknown` return 0
|
|
- Existing `test_lookup_model_context_prefix_match` still passes (deepseek-r1:7b and llama-3.1-8b-instruct)
|