82e14e7ec8
--fa4_prefill now accepts three states: True (force on, warn + fall back on sub-Hopper), False (force off), None / default (auto-enable where supported). Argparse switches to BooleanOptionalAction so both --fa4_prefill and --no-fa4_prefill work, with the default being auto-detect from torch.cuda.get_device_capability. Adds a cu13 / cu12 install section and a per-GPU support matrix to scripts/benchmarks/README.md. Adds tests/test_fa4_capability_guard.py covering the nine combinations of (explicit-on / auto / explicit-off) x (sm_80 / sm_90 / sm_100 / sm_120). Monkey-patches get_device_capability and stubs PageTable / patch_qwen3_model so it runs without CUDA. |
||
|---|---|---|
| .. | ||
| python | ||
| qlora | ||
| saving | ||
| sh | ||
| studio/install | ||
| utils | ||
| __init__.py | ||
| run_all.sh | ||
| test_cli_export_unpacking.py | ||
| test_fa4_capability_guard.py | ||
| test_get_model_name.py | ||
| test_loader_glob_skip.py | ||
| test_model_registry.py | ||
| test_raw_text.py | ||