tlg 813bbe0ad0 fix: VRAM eviction cascades through all tiers for large LLM loads ...

The original eviction logic blocked ASR eviction even when an LLM
genuinely needed all 16GB VRAM (e.g., gpt-oss-20b at 13GB). Now uses
two-pass eviction: first evicts lower/same priority, then cascades to
higher priority as last resort. Added tests for ASR-survives and
full-cascade scenarios.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

2026-04-04 09:22:14 +02:00

..

__init__.py

feat: project scaffolding with config files and test fixtures

2026-04-04 07:23:14 +02:00

conftest.py

feat: project scaffolding with config files and test fixtures

2026-04-04 07:23:14 +02:00

test_auth.py

feat: API key authentication dependency

2026-04-04 07:31:30 +02:00

test_config.py

feat: config loading for models.yaml and api_keys.yaml

2026-04-04 07:30:13 +02:00

test_model_registry.py

feat: model registry with virtual-to-physical resolution

2026-04-04 07:31:10 +02:00

test_vram_manager.py

fix: VRAM eviction cascades through all tiers for large LLM loads

2026-04-04 09:22:14 +02:00