loopyt ec5f37821d fix(h20): let the reference backend sync, and stop the reference raising under the guard ... Running end to end surfaced two things the per-kernel A/B could not. The engine wraps the forward pass in `forbid_cuda_sync()`, which sets torch's sync debug mode to "error". The compiled kernels never sync; the torch references do, and not by oversight: * `tile_kernels/torch/cast.py` built `torch.tensor(-1.0, device="cuda")` (and three more like it) for use as a `torch.where` operand. Constructing a device tensor from a Python scalar is a host-to-device copy, so it syncs -- and it is unnecessary, because `torch.where` accepts a plain scalar and passes it to the kernel. Patched via `patch_torch_ref_sync.py` (idempotent, reversible), which is a fix worth having on its own: the same cast raises for any caller that holds the device clean. * `get_mapping_from_psum` reads the expert prefix-sum to the host to compute alignment ranges in Python, and `moe_topk_gate_forward` compacts rows with `nonzero()` and branches on `.any()`. Those cannot be made sync-free without rewriting how the reference addresses, which is the wrong trade for a fallback whose job is being runnable. So each compat module wraps its reference module once and every call through it runs inside `allow_cuda_sync()`. Not done, deliberately: `set_sync_debug_mode("default")` process-wide. That would also stop the guard catching accidental syncs in the compiled path, which is what it is for. Also fixed: my own `bool(force_random.any())` in the gate adapter was itself a sync, outside the reference, so it raised before the adapter could decide anything. End to end now completes: exit=0, all 8 DP ranks agree, coherent answers. It is not token-identical to the compiled path -- 16/16 samples differ, with a control run confirming the harness is deterministic (two cuda runs are bit-identical), so the difference is real. Recorded in the doc as the next thing to attribute rather than papered over.		2026-10-24 19:10:49 +08:00
..
compare_with_cuda.py	feat(h20): make the pure-torch kernel references usable without TileLang	2026-10-22 16:08:46 +08:00
compat_ab.py	feat(h20): route tile_kernels calls through a compat layer, so the model runs without TileLang	2026-10-22 18:10:14 +08:00
lazify_package_init.py	feat(h20): make the pure-torch kernel references usable without TileLang	2026-10-22 16:08:46 +08:00
patch_quant_common.py	feat(h20): make the pure-torch kernel references usable without TileLang	2026-10-22 16:08:46 +08:00
patch_tile_kernels.py	feat(h20): make the pure-torch kernel references usable without TileLang	2026-10-22 16:08:46 +08:00
patch_torch_ref_sync.py	fix(h20): let the reference backend sync, and stop the reference raising under the guard	2026-10-24 19:10:49 +08:00
replace_imports.py	feat(h20): route tile_kernels calls through a compat layer, so the model runs without TileLang	2026-10-22 18:10:14 +08:00
scan_compat.py	feat(h20): route tile_kernels calls through a compat layer, so the model runs without TileLang	2026-10-22 18:10:14 +08:00
tk_patch_init.py	feat(h20): make the pure-torch kernel references usable without TileLang	2026-10-22 16:08:46 +08:00