miniworld-engine 2.3.10__tar.gz → 2.3.12__tar.gz

{miniworld_engine-2.3.10 → miniworld_engine-2.3.12}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: miniworld-engine
-Version: 2.3.10
+Version: 2.3.12
 Summary: MiniWorld GPU kernel development repo — Triton / CUDA kernels + benches for AF3-style ops (triangle multiplication, transition, attention, ...).
 License-Expression: MIT AND Apache-2.0 AND BSD-3-Clause
 Requires-Python: >=3.10
@@ -39,7 +39,7 @@ Dynamic: license-file
 # miniworld-engine
 
 GPU kernel library for MiniWorld / AF3-style ops. Each op is cut out of the model and
-optimized in isolation: **hand-written CUDA where it exists, a Triton fallback on every other GPU**
+optimized in isolation: **hand-written CUDA where it exists, a portable kernel fallback on every other GPU**
 (H100 and B200 have none, on purpose: they fall back to PyTorch), and a PyTorch reference that defines what "correct" means.
 
 > **Where this fits.** miniworld-engine is the bottom layer of a three-layer stack: it owns
@@ -47,7 +47,7 @@ optimized in isolation: **hand-written CUDA where it exists, a Triton fallback o
 > blocks; terminal product repos assemble those into full models. Boundary rules live in
 > team-gm's `docs/ARCHITECTURE.md`.
 
-**Version 2.3.10** — CUDA + Triton only (CuTe/quack removed), torch 2.13 (CUDA 12.9 or 13.0), prebuilt H100 kernels,
+**Version 2.3.12** — CUDA + portable kernels only (CuTe/quack removed), torch 2.13 (CUDA 12.9 or 13.0), prebuilt H100 kernels,
 cuequivariance 0.12. [Changelog](docs/CHANGELOG.md) ·
 [attribution](licenses/THIRD_PARTY_NOTICES.md).
 
@@ -71,7 +71,7 @@ to read a number.
 python -m pip install -e .
 ```
 
-**2. Check what you have.** Prints the version and the config set every triton kernel will search
+**2. Check what you have.** Prints the version and the config set every kernel will search
 if you do nothing else.
 
 ```bash
@@ -199,12 +199,12 @@ from the code.
 
 **The minimum supported architecture is sm80.** The generated table above states each
 registered kernel's minimum architecture; it is not a numerical qualification for every
-listed GPU. The public backends are `pytorch`, `miniworld`, `cuequivariance` and `anthropic`; Triton is
-not one of them. `miniworld` runs the engine's CUDA path where one serves the call; on a card without one
-it routes to the internal Triton kernel family only if the engine holds a record for that exact card
+listed GPU. The values of `implementation` are `pytorch`, `miniworld`, `cuequivariance` and `anthropic`; `miniworld` routes per card
+automatically. It runs the engine's CUDA path where one serves the call; on a card without one
+it routes to the engine's portable kernels only if the engine holds a record for that exact card
 (`autotune/manifests/`), else it warns and runs the PyTorch reference
-([docs/dispatch.md](docs/dispatch.md)). `settings.force_triton_kernels` is the internal A/B switch that
-forces the Triton family. Native CUDA variants can require a newer architecture.
+([docs/dispatch.md](docs/dispatch.md)). `settings.force_portable_kernels` is the internal A/B switch that
+forces the portable kernels. Native CUDA variants can require a newer architecture.
 
 One extension is **not** in the table because it is not in the registry: `transition_b2b_cuda`,
 which the `Transition` module builds on demand, is compiled for `sm_90a` and fails to build on

{miniworld_engine-2.3.10 → miniworld_engine-2.3.12}/README.md

@@ -1,7 +1,7 @@
 # miniworld-engine
 
 GPU kernel library for MiniWorld / AF3-style ops. Each op is cut out of the model and
-optimized in isolation: **hand-written CUDA where it exists, a Triton fallback on every other GPU**
+optimized in isolation: **hand-written CUDA where it exists, a portable kernel fallback on every other GPU**
 (H100 and B200 have none, on purpose: they fall back to PyTorch), and a PyTorch reference that defines what "correct" means.
 
 > **Where this fits.** miniworld-engine is the bottom layer of a three-layer stack: it owns
@@ -9,7 +9,7 @@ optimized in isolation: **hand-written CUDA where it exists, a Triton fallback o
 > blocks; terminal product repos assemble those into full models. Boundary rules live in
 > team-gm's `docs/ARCHITECTURE.md`.
 
-**Version 2.3.10** — CUDA + Triton only (CuTe/quack removed), torch 2.13 (CUDA 12.9 or 13.0), prebuilt H100 kernels,
+**Version 2.3.12** — CUDA + portable kernels only (CuTe/quack removed), torch 2.13 (CUDA 12.9 or 13.0), prebuilt H100 kernels,
 cuequivariance 0.12. [Changelog](docs/CHANGELOG.md) ·
 [attribution](licenses/THIRD_PARTY_NOTICES.md).
 
@@ -33,7 +33,7 @@ to read a number.
 python -m pip install -e .
 ```
 
-**2. Check what you have.** Prints the version and the config set every triton kernel will search
+**2. Check what you have.** Prints the version and the config set every kernel will search
 if you do nothing else.
 
 ```bash
@@ -161,12 +161,12 @@ from the code.
 
 **The minimum supported architecture is sm80.** The generated table above states each
 registered kernel's minimum architecture; it is not a numerical qualification for every
-listed GPU. The public backends are `pytorch`, `miniworld`, `cuequivariance` and `anthropic`; Triton is
-not one of them. `miniworld` runs the engine's CUDA path where one serves the call; on a card without one
-it routes to the internal Triton kernel family only if the engine holds a record for that exact card
+listed GPU. The values of `implementation` are `pytorch`, `miniworld`, `cuequivariance` and `anthropic`; `miniworld` routes per card
+automatically. It runs the engine's CUDA path where one serves the call; on a card without one
+it routes to the engine's portable kernels only if the engine holds a record for that exact card
 (`autotune/manifests/`), else it warns and runs the PyTorch reference
-([docs/dispatch.md](docs/dispatch.md)). `settings.force_triton_kernels` is the internal A/B switch that
-forces the Triton family. Native CUDA variants can require a newer architecture.
+([docs/dispatch.md](docs/dispatch.md)). `settings.force_portable_kernels` is the internal A/B switch that
+forces the portable kernels. Native CUDA variants can require a newer architecture.
 
 One extension is **not** in the table because it is not in the registry: `transition_b2b_cuda`,
 which the `Transition` module builds on demand, is compiled for `sm_90a` and fails to build on

{miniworld_engine-2.3.10 → miniworld_engine-2.3.12}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "miniworld-engine"
-version = "2.3.10"
+version = "2.3.12"
 requires-python = ">=3.10"
 readme = "README.md"
 description = "MiniWorld GPU kernel development repo — Triton / CUDA kernels + benches for AF3-style ops (triangle multiplication, transition, attention, ...)."

{miniworld_engine-2.3.10 → miniworld_engine-2.3.12}/src/miniworld_engine/architecture/_engine_impl.py

@@ -16,7 +16,7 @@ def torch_pointwise() -> bool:
 
     B200 policy: the pieces around the GEMMs and attention are PyTorch ops that
     ``torch.compile`` fuses (inductor), not the engine's hand-written Triton
-    kernels. ``miniworld_engine.settings.b200_engine_triton`` keeps the engine
+    kernels. ``miniworld_engine.settings.b200_portable_kernels`` keeps the engine
     Triton kernels (the A/B switch). An engine that predates the switch keeps
     its Triton kernels, so this is a no-op until the engine pin has it.
 
@@ -28,7 +28,7 @@ def torch_pointwise() -> bool:
         and torch.cuda.get_device_capability() == _B200_CAPABILITY
     ):
         return False
-    return not getattr(settings.current(), "b200_engine_triton", True)
+    return not getattr(settings.current(), "b200_portable_kernels", True)
 
 
 def to_engine_impl_pointwise(

{miniworld_engine-2.3.10 → miniworld_engine-2.3.12}/src/miniworld_engine/autotune/builder.py

@@ -146,7 +146,7 @@ SWITCHES: dict[str, tuple[tuple, tuple[str, ...]]] = {
     # covered by the unpinned unit, so pinning both would double the build for nothing.
     "ln_bwd_path": (("persistent", "atomic", "cuda"), ("train",)),
     "ln_out_bwd_path": (("split", "fused"), ("train",)),
-    "trimul_impl": (("triton",), ("eval", "train")),
+    "trimul_impl": (("portable",), ("eval", "train")),
     # The rest of settings.py's backend selectors. They were left out when SWITCHES was written
     # and each one is a set of kernels no unit ever reached: measured against the shipped A6000
     # cache, 26 kernels with working drivers were in the cache and in NO derived unit, and the
@@ -318,8 +318,8 @@ def cases() -> list[Case]:
     )
 
     def IT(i):
-        # "triton" is the A/B label of a build unit (run_case forces the Triton family around it), not a public backend
-        return ImplementationType.MINIWORLD if i == "triton" else ImplementationType(i)
+        # "portable" is the A/B label of a build unit (run_case forces the portable family around it), not a backend
+        return ImplementationType.MINIWORLD if i == "portable" else ImplementationType(i)
 
     # The dims that used to be declared here -- PAIR_D, PAIR_HID, HID_D, BOTH -- are in
     # registry_module.csv now, with a `source` column saying where each number came from. They are
@@ -464,11 +464,11 @@ def cases() -> list[Case]:
 def run_case(case: Case, length: int, dim_index: int, *, train: bool, p_drop: float = 0.0,
              impl: str = "miniworld", dtype: torch.dtype = BF16,
              compute_dtype: torch.dtype | None = None) -> int:
-    """:func:`_run_case`, with the engine's Triton family forced for the ``triton`` impl: it is not a public backend any more, so the unit
-    builds a ``miniworld`` module under ``settings.force_triton_kernels`` (the A/B switch), as the explicit ``triton`` request did."""
+    """:func:`_run_case`, with the portable family forced for the ``portable`` impl: it is not a backend, so the unit
+    builds a ``miniworld`` module under ``settings.force_portable_kernels`` (the A/B switch), as the A/B request does."""
     from miniworld_engine import settings
 
-    previous = settings.configure(force_triton_kernels=True) if impl == "triton" else None
+    previous = settings.configure(force_portable_kernels=True) if impl == "portable" else None
     try:
         return _run_case(case, length, dim_index, train=train, p_drop=p_drop, impl=impl, dtype=dtype, compute_dtype=compute_dtype)
     finally:

{miniworld_engine-2.3.10 → miniworld_engine-2.3.12}/src/miniworld_engine/autotune/manifests/NVIDIA H100 80GB HBM3 (sm90).csv

@@ -1,5 +1,5 @@
 kernel,backend,family,file,status,detail,dtype
-#provenance,2.3.9,d6d1eb084d32,clean,2026-10-24,v2.3.8-9-gd6d1eb08,
+#provenance,2.3.11,e5a9bd768e0f,clean,2026-10-24,v2.3.10-10-ge5a9bd76,
 adaln_bwd_dx_dlnw_triton,triton,adaln,miniworld_engine/kernels/adaln/triton/training.py,ok,rel DCond=2.15e-03 DLnW=2.62e-03 (band 1e-02 declared),bf16
 adaln_bwd_pre_dx_triton,triton,adaln,miniworld_engine/kernels/adaln/triton/training.py,ok,rel D=1.49e-03 DX=3.13e-03 (band 1e-02 declared),bf16
 adaln_epilogue_saveact_triton,triton,adaln,miniworld_engine/kernels/adaln/triton/training.py,ok,rel Y=2.65e-03 Mean=2.23e-08 Rstd=1.86e-07 Gate=1.96e-03 (band 1e-02 declared),bf16
@@ -11,6 +11,9 @@ augmented_attention_bwd_atomic_triton,triton,augmented_attention,miniworld_engin
 augmented_attention_bwd_pre_triton,triton,augmented_attention,miniworld_engine/kernels/augmented_attention/triton/main.py,ok,rel out=8.02e-08 (band 4e-07 declared),bf16
 augmented_attention_bwd_split_triton,triton,augmented_attention,miniworld_engine/kernels/augmented_attention/triton/main.py,ok,rel dq=4.04e-03 dk=3.25e-03 dv=4.70e-03 dbias=2.58e-03 (band 2e-02 declared),bf16
 augmented_attention_fwd_triton,triton,augmented_attention,miniworld_engine/kernels/augmented_attention/triton/main.py,ok,rel out=2.17e-03 m=9.38e-08 (band 9e-03 declared),bf16
+augmented_attention_fwd_window_triton,triton,augmented_attention,miniworld_engine/kernels/augmented_attention/triton/window.py,ok,rel out=2.29e-03 lse=1.23e-07 (band 1e-02 declared),bf16
+augmented_attention_bwd_window_triton,triton,augmented_attention,miniworld_engine/kernels/augmented_attention/triton/window.py,ok,rel dq=3.33e-03 dk=3.39e-03 dv=2.44e-03 dbias=2.12e-03 (band 1e-02 declared),bf16
+augmented_attention_bwd_reduce_window_triton,triton,augmented_attention,miniworld_engine/kernels/augmented_attention/triton/window.py,ok,rel out=8.28e-08 (band 3e-07 declared),bf16
 bias_only_attention_bwd_pre_triton,triton,bias_only_attention,miniworld_engine/kernels/bias_only_attention/triton/main.py,ok,rel out=6.62e-08 (band 3e-07 declared),bf16
 bias_only_attention_bwd_triton,triton,bias_only_attention,miniworld_engine/kernels/bias_only_attention/triton/main.py,ok,rel dv=7.22e-03 dbias=5.84e-03 (band 5e-02 declared),bf16
 bias_only_attention_fwd_triton,triton,bias_only_attention,miniworld_engine/kernels/bias_only_attention/triton/main.py,ok,rel out=8.88e-03 m=8.11e-04 (band 4e-02 declared),bf16
@@ -32,7 +35,7 @@ gated_projection_gate_dropres_triton,triton,trimul_inproj,miniworld_engine/kerne
 gated_projection_gate_flat_triton,triton,gated_projection,miniworld_engine/kernels/gated_projection/triton/main.py,ok,rel out=2.15e-03 (band 8e-03 declared),bf16
 gated_projection_gate_gemm_triton,triton,bias_only_attention,miniworld_engine/kernels/bias_only_attention/triton/gate_out.py,ok,rel out=3.79e-03 (band 1e-02 declared),bf16
 gated_projection_gate_triton,triton,gated_projection,miniworld_engine/kernels/gated_projection/triton/main.py,ok,rel out=2.15e-03 (band 8e-03 declared),bf16
-layernorm_bwd_atomic_strided_triton,triton,layernorm_linear,miniworld_engine/kernels/layernorm_linear/triton/te_style.py,ok,rel dx=2.60e-03 dgamma=3.78e-07 dbeta=3.34e-07 (band 1e-02 declared),bf16
+layernorm_bwd_atomic_strided_triton,triton,layernorm_linear,miniworld_engine/kernels/layernorm_linear/triton/te_style.py,ok,rel dx=2.60e-03 dgamma=5.35e-07 dbeta=1.25e-07 (band 1e-02 declared),bf16
 layernorm_bwd_atomic_triton,triton,layernorm,miniworld_engine/kernels/layernorm/triton/main.py,ok,rel dx=2.03e-03 dw=2.26e-03 db=2.41e-03 (band 1e-02 declared),bf16
 layernorm_bwd_foldstats_triton,triton,transition,miniworld_engine/kernels/transition/triton/fused.py,ok,rel dx=2.61e-03 dgamma=2.40e-07 dbeta=8.45e-08 (band 9e-03 declared),bf16
 layernorm_bwd_reduce_cuda,cuda,layernorm,miniworld_engine/kernels/layernorm/cuda/layer_norm_cuda_kernel.cu,ok,rel dw=2.26e-03 db=2.41e-03 (band 9e-03 declared),bf16
@@ -49,25 +52,28 @@ layernorm_fwd_strided_triton,triton,adaln,miniworld_engine/kernels/adaln/triton/
 layernorm_linear_bwd_fp32_triton,triton,layernorm_linear,miniworld_engine/kernels/layernorm_linear/triton/pair_bias.py,ok,rel dx=2.30e-03 dln_weight=1.74e-03 dproj_weight=2.01e-03 (band 1e-02 declared),bf16
 layernorm_linear_fwd_fp32_triton,triton,layernorm_linear,miniworld_engine/kernels/layernorm_linear/triton/pair_bias.py,ok,rel out=2.37e-03 mean=2.08e-08 rstd=1.74e-07 (band 9e-03 declared),bf16
 layernorm_linear_fwd_triton,triton,layernorm_linear,miniworld_engine/kernels/layernorm_linear/triton/fused.py,ok,rel out=3.51e-03 (band 2e-02 declared),bf16
-layernorm_stats_triton,triton,layernorm_linear,miniworld_engine/kernels/layernorm_linear/triton/stats.py,ok,rel rstd=8.69e-08 c1=8.17e-08 (band 6e-07 declared),bf16
+layernorm_stats_triton,triton,layernorm_linear,miniworld_engine/kernels/layernorm_linear/triton/stats.py,ok,rel rstd=8.69e-08 c1=1.63e-07 (band 6e-07 declared),bf16
 outer_product_mean_bwd_epilogue_dw_triton,triton,outer_product_mean,miniworld_engine/kernels/outer_product_mean/triton/main.py,ok,rel dw_out=4.55e-03 (band 1e-02 declared),bf16
 outer_product_mean_bwd_epilogue_dx_triton,triton,outer_product_mean,miniworld_engine/kernels/outer_product_mean/triton/main.py,ok,rel dmsa=5.84e-03 db_out=2.05e-03 dresidual=0.00e+00 (band 1e-02 declared),bf16
-outer_product_mean_bwd_layernorm_gemm_triton,triton,outer_product_mean,miniworld_engine/kernels/outer_product_mean/triton/main.py,ok,rel dmsa=5.84e-03 dln_weight=4.26e-03 dln_bias=3.54e-03 dw_left=4.97e-03 dw_right=6.01e-03 (band 1e-02 declared),bf16
+outer_product_mean_bwd_layernorm_gemm_triton,triton,outer_product_mean,miniworld_engine/kernels/outer_product_mean/triton/main.py,ok,rel dmsa=5.84e-03 dln_weight=4.26e-03 dln_bias=3.54e-03 dw_left=4.97e-03 dw_right=6.01e-03 dmsa_pbias=4.23e-03 dln_weight_pbias=7.72e-03 dln_bias_pbias=2.32e-03 dw_left_pbias=5.16e-03 dw_right_pbias=4.23e-03 db_left_pbias=1.97e-03 db_right_pbias=3.41e-03 dmsa_grouped=4.30e-03 dln_weight_grouped=4.27e-03 dln_bias_grouped=2.00e-03 dw_left_grouped=5.39e-03 dw_right_grouped=4.50e-03 (band 1e-02 declared),bf16
 outer_product_mean_epilogue_triton,triton,outer_product_mean,miniworld_engine/kernels/outer_product_mean/triton/main.py,ok,rel out=2.88e-03 (band 1e-02 declared),bf16
-outer_product_mean_layernorm_gemm_triton,triton,outer_product_mean,miniworld_engine/kernels/outer_product_mean/triton/main.py,ok,rel out=2.88e-03 (band 1e-02 declared),bf16
-pair_weighted_averaging_bwd_gate_gemm_triton,triton,pair_weighted_averaging,miniworld_engine/kernels/pair_weighted_averaging/triton/main.py,ok,rel dmsa=2.84e-03 dw_gate=4.47e-03 dw_out=3.30e-03 (band 1e-02 declared),bf16
-pair_weighted_averaging_bwd_layernorm_gemm_dw_triton,triton,pair_weighted_averaging,miniworld_engine/kernels/pair_weighted_averaging/triton/main.py,ok,rel dw_value=4.19e-03 (band 1e-02 declared),bf16
-pair_weighted_averaging_bwd_layernorm_gemm_dx_dlnw_triton,triton,pair_weighted_averaging,miniworld_engine/kernels/pair_weighted_averaging/triton/main.py,ok,rel dmsa=2.84e-03 dln_msa_weight=4.39e-03 dln_msa_bias=4.52e-03 (band 1e-02 declared),bf16
-pair_weighted_averaging_bwd_layernorm_gemm_softmax_triton,triton,pair_weighted_averaging,miniworld_engine/kernels/pair_weighted_averaging/triton/main.py,ok,rel dpair=4.99e-03 dln_pair_weight=3.87e-03 dw_bias=4.26e-03 (band 1e-02 declared),bf16
-pair_weighted_averaging_gate_gemm_triton,triton,pair_weighted_averaging,miniworld_engine/kernels/pair_weighted_averaging/triton/main.py,ok,rel out=3.02e-03 (band 1e-02 declared),bf16
-pair_weighted_averaging_layernorm_gemm_softmax_triton,triton,pair_weighted_averaging,miniworld_engine/kernels/pair_weighted_averaging/triton/main.py,ok,rel out=3.02e-03 (band 1e-02 declared),bf16
-pair_weighted_averaging_layernorm_gemm_triton,triton,pair_weighted_averaging,miniworld_engine/kernels/pair_weighted_averaging/triton/main.py,ok,rel out=3.02e-03 (band 1e-02 declared),bf16
+outer_product_mean_bwd_epilogue_ln_dw_triton,triton,outer_product_mean,miniworld_engine/kernels/outer_product_mean/triton/main.py,ok,rel dw_out_grouped=4.30e-03 (band 2e-02 declared),bf16
+outer_product_mean_bwd_epilogue_ln_dx_triton,triton,outer_product_mean,miniworld_engine/kernels/outer_product_mean/triton/main.py,ok,rel dmsa_grouped=4.71e-03 db_out_grouped=2.05e-03 dresidual_grouped=0.00e+00 dproduct_weight_grouped=6.06e-03 dproduct_bias_grouped=2.72e-03 (band 3e-02 declared),bf16
+outer_product_mean_epilogue_ln_triton,triton,outer_product_mean,miniworld_engine/kernels/outer_product_mean/triton/main.py,ok,rel out_grouped=5.06e-03 (band 2e-02 declared),bf16
+outer_product_mean_layernorm_gemm_triton,triton,outer_product_mean,miniworld_engine/kernels/outer_product_mean/triton/main.py,ok,rel out=2.88e-03 out_pbias=2.82e-03 out_grouped=5.06e-03 (band 1e-02 declared),bf16
+pair_weighted_averaging_bwd_gate_gemm_triton,triton,pair_weighted_averaging,miniworld_engine/kernels/pair_weighted_averaging/triton/main.py,ok,rel dmsa=2.84e-03 dw_gate=4.47e-03 dw_out=3.30e-03 dmsa_c8=2.69e-03 dw_gate_c8=4.03e-03 dw_out_c8=4.06e-03 (band 1e-02 declared),bf16
+pair_weighted_averaging_bwd_layernorm_gemm_dw_triton,triton,pair_weighted_averaging,miniworld_engine/kernels/pair_weighted_averaging/triton/main.py,ok,rel dw_value=4.19e-03 dw_value_c8=3.40e-03 (band 1e-02 declared),bf16
+pair_weighted_averaging_bwd_layernorm_gemm_dx_dlnw_triton,triton,pair_weighted_averaging,miniworld_engine/kernels/pair_weighted_averaging/triton/main.py,ok,rel dmsa=2.84e-03 dln_msa_weight=4.39e-03 dln_msa_bias=4.52e-03 dmsa_c8=2.69e-03 dln_msa_weight_c8=5.20e-03 dln_msa_bias_c8=2.51e-03 (band 1e-02 declared),bf16
+pair_weighted_averaging_bwd_layernorm_gemm_softmax_triton,triton,pair_weighted_averaging,miniworld_engine/kernels/pair_weighted_averaging/triton/main.py,ok,rel dpair=4.99e-03 dln_pair_weight=3.87e-03 dw_bias=4.26e-03 dpair_c8=5.66e-03 dln_pair_weight_c8=6.34e-03 dw_bias_c8=7.50e-03 (band 1e-02 declared),bf16
+pair_weighted_averaging_gate_gemm_triton,triton,pair_weighted_averaging,miniworld_engine/kernels/pair_weighted_averaging/triton/main.py,ok,rel out=3.02e-03 out_c8=2.99e-03 (band 1e-02 declared),bf16
+pair_weighted_averaging_layernorm_gemm_softmax_triton,triton,pair_weighted_averaging,miniworld_engine/kernels/pair_weighted_averaging/triton/main.py,ok,rel out=3.02e-03 out_c8=2.99e-03 (band 1e-02 declared),bf16
+pair_weighted_averaging_layernorm_gemm_triton,triton,pair_weighted_averaging,miniworld_engine/kernels/pair_weighted_averaging/triton/main.py,ok,rel out=3.02e-03 out_c8=2.99e-03 (band 1e-02 declared),bf16
 rmsnorm_adamod_bwd_triton,triton,rmsnorm_adamod,miniworld_engine/kernels/rmsnorm/triton/main.py,ok,rel dq_plain=2.56e-03 dc_plain=4.44e-03 dWsc_plain=4.10e-03 dWsh_plain=1.76e-03 dWg_plain=4.17e-03 dq_aff=1.55e-03 dc_aff=4.22e-03 dWsc_aff=4.48e-03 dWsh_aff=1.76e-03 dWg_aff=4.17e-03 dweight_aff=0.00e+00 (band 1e-02 declared),bf16
 rmsnorm_adamod_fwd_triton,triton,rmsnorm_adamod,miniworld_engine/kernels/rmsnorm/triton/main.py,ok,rel y_plain=2.89e-03 gate_plain=1.94e-03 y_aff=1.34e-03 gate_aff=1.94e-03 (band 5e-03 declared),bf16
 rmsnorm_bwd_triton,triton,rmsnorm,miniworld_engine/kernels/rmsnorm/triton/main.py,ok,rel dx_aff=2.03e-03 dweight_aff=2.73e-03 dx_plain=2.54e-03 (band 4e-03 declared),bf16
 rmsnorm_fwd_triton,triton,rmsnorm,miniworld_engine/kernels/rmsnorm/triton/main.py,ok,rel y_aff=2.37e-03 y_plain=3.80e-03 (band 4e-03 declared),bf16
 rope_fwd_triton,triton,rope,miniworld_engine/kernels/rope/triton/main.py,ok,rel y=5.14e-04 dx=7.23e-04 (band 3e-03 declared),bf16
-transition_bwd_cuda,cuda,transition,miniworld_engine/kernels/transition/cuda/transition_cuda_kernel.cu,failed,RuntimeError: Error building extension 'transition_cuda_ext_v2': [1/2] c++ -MMD -MF transition_cuda.o.d -DTORCH_EXTENSION_NAME=transit | c++: error: unrecognized command line option ‘-std=c++20’; did you mean ‘-std=c++2a’?,bf16
+transition_bwd_cuda,cuda,transition,miniworld_engine/kernels/transition/cuda/transition_cuda_kernel.cu,failed,RuntimeError: Error building extension 'transition_cuda_ext_v2': [1/3] c++ -MMD -MF transition_cuda.o.d -DTORCH_EXTENSION_NAME=transit | c++: error: unrecognized command line option ‘-std=c++20’; did you mean ‘-std=c++2a’?,bf16
 transition_bwd_swiglu_recompute_triton,triton,transition,miniworld_engine/kernels/transition/triton/fused.py,ok,rel h=2.18e-03 dAB=2.12e-03 xn=7.02e-04 (band 1e-02 declared),bf16
 transition_cast_cuda,cuda,transition,miniworld_engine/kernels/transition/cuda/transition_cuda_kernel.cu,failed,ImportError: <path>,bf16
 transition_expand_swiglu_triton,triton,transition,miniworld_engine/kernels/transition/triton/main.py,ok,rel out=2.81e-03 (band 1e-02 declared),bf16
@@ -127,22 +133,22 @@ mpnn_edge_dropout_bwd_packed_triton,triton,mpnn_edge_dropout,miniworld_engine/ke
 trimul_fwd_sm90_cuda,cuda,trimul_inproj,miniworld_engine/kernels/trimul_inproj/cuda/h100_inference.py,ok,rel h128_dir0=4.58e-03 h128_dir1=4.64e-03 h128_dir2=4.63e-03 h256_dir0=5.06e-03 h256_dir1=4.88e-03 h256_dir2=5.34e-03 (band 2e-02 declared),bf16
 transition_fwd_residual_sm90_cuda,cuda,transition,miniworld_engine/kernels/transition/cuda/transition_fused_fwd_sm90a_kernel.cu,ok,rel n4_y=3.57e-03 n4_xn=3.10e-03 n4_rstd=8.69e-08 n4_c1=6.69e-08 n2_y=3.33e-03 n2_xn=3.10e-03 n2_rstd=8.69e-08 n2_c1=6.69e-08 (band 1e-02 declared),bf16
 transition_bwd_residual_sm90_cuda,cuda,transition,miniworld_engine/kernels/transition/cuda/transition_fused_bwd_sm90a_kernel.cu,ok,rel n4_dx=5.85e-03 n4_dgamma=4.31e-03 n4_dbeta=3.03e-03 n4_dWa=3.97e-03 n4_dWb=4.61e-03 n4_dWs=4.17e-03 n2_dx=6.01e-03 n2_dgamma=4.39e-03 n2_dbeta=3.81e-03 n2_dWa=3.92e-03 n2_dWb=4.27e-03 n2_dWs=3.68e-03 (band 2e-02 declared),bf16
-swa_dit_inproj_fwd_triton,triton,swa_dit,miniworld_engine/kernels/swa_dit/triton/forward.py,ok,rel out=5.94e-03 (band 5e-02 declared),bf16
-swa_dit_softmax_fwd_triton,triton,swa_dit,miniworld_engine/kernels/swa_dit/triton/forward.py,ok,rel out=5.94e-03 (band 5e-02 declared),bf16
-swa_dit_output_swiglu_fwd_triton,triton,swa_dit,miniworld_engine/kernels/swa_dit/triton/forward.py,ok,rel out=5.94e-03 (band 5e-02 declared),bf16
+swa_dit_inproj_fwd_triton,triton,swa_dit,miniworld_engine/kernels/swa_dit/triton/forward.py,ok,rel out=5.90e-03 (band 5e-02 declared),bf16
+swa_dit_softmax_fwd_triton,triton,swa_dit,miniworld_engine/kernels/swa_dit/triton/forward.py,ok,rel out=5.90e-03 (band 5e-02 declared),bf16
+swa_dit_output_swiglu_fwd_triton,triton,swa_dit,miniworld_engine/kernels/swa_dit/triton/forward.py,ok,rel out=5.90e-03 (band 5e-02 declared),bf16
 swa_dit_swiglu_bwd_triton,triton,swa_dit,miniworld_engine/kernels/swa_dit/triton/backward.py,ok,rel dmod=3.40e-03 dw_up=4.71e-03 dw_down=4.70e-03 (band 5e-02 declared),bf16
 swa_dit_swiglu_dw_triton,triton,swa_dit,miniworld_engine/kernels/swa_dit/triton/backward.py,ok,rel dw_up=4.71e-03 dw_down=4.70e-03 (band 5e-02 declared),bf16
-swa_dit_output_bwd_triton,triton,swa_dit,miniworld_engine/kernels/swa_dit/triton/backward.py,ok,rel dmod=3.40e-03 dw_out=4.66e-03 (band 5e-02 declared),bf16
-swa_dit_softmax_bwd_dq_triton,triton,swa_dit,miniworld_engine/kernels/swa_dit/triton/backward.py,ok,rel dq=6.18e-03 dw_qkv=5.72e-03 (band 5e-02 declared),bf16
-swa_dit_softmax_bwd_dkdv_triton,triton,swa_dit,miniworld_engine/kernels/swa_dit/triton/backward.py,ok,rel dq=6.18e-03 dw_qkv=5.72e-03 (band 5e-02 declared),bf16
-swa_dit_inproj_bwd_triton,triton,swa_dit,miniworld_engine/kernels/swa_dit/triton/backward.py,ok,rel dq=6.18e-03 dmod=3.40e-03 dw_qkv=5.72e-03 dw_gate=4.68e-03 (band 5e-02 declared),bf16
+swa_dit_output_bwd_triton,triton,swa_dit,miniworld_engine/kernels/swa_dit/triton/backward.py,ok,rel dmod=3.40e-03 dw_out=4.32e-03 (band 5e-02 declared),bf16
+swa_dit_softmax_bwd_dq_triton,triton,swa_dit,miniworld_engine/kernels/swa_dit/triton/backward.py,ok,rel dq=6.18e-03 dw_qkv=6.74e-03 (band 5e-02 declared),bf16
+swa_dit_softmax_bwd_dkdv_triton,triton,swa_dit,miniworld_engine/kernels/swa_dit/triton/backward.py,ok,rel dq=6.18e-03 dw_qkv=6.74e-03 (band 5e-02 declared),bf16
+swa_dit_inproj_bwd_triton,triton,swa_dit,miniworld_engine/kernels/swa_dit/triton/backward.py,ok,rel dq=6.18e-03 dmod=3.40e-03 dw_qkv=6.74e-03 dw_gate=4.86e-03 (band 5e-02 declared),bf16
 swa_dit_inproj_fwd_sm90_cuda,cuda,swa_dit,miniworld_engine/kernels/swa_dit/cuda/swa_qkvg_fwd.cu,ok,rel out=5.90e-03 (band 5e-02 declared),bf16
-swa_dit_output_swiglu_fwd_sm90_cuda,cuda,swa_dit,miniworld_engine/kernels/swa_dit/cuda/swa_ffn_fwd.cu,ok,rel out=5.94e-03 (band 5e-02 declared),bf16
+swa_dit_output_swiglu_fwd_sm90_cuda,cuda,swa_dit,miniworld_engine/kernels/swa_dit/cuda/swa_ffn_fwd.cu,ok,rel out=5.90e-03 (band 5e-02 declared),bf16
 swa_dit_swiglu_bwd_sm90_cuda,cuda,swa_dit,miniworld_engine/kernels/swa_dit/cuda/swa_ffn_bwd.cu,ok,rel dq=6.18e-03 dmod=3.40e-03 dw_up=4.71e-03 dw_down=4.70e-03 (band 5e-02 declared),bf16
 swa_dit_softmax_fwd_sm90_cuda,cuda,swa_dit,miniworld_engine/kernels/swa_dit/cuda/swa_attn.cu,ok,rel out=5.90e-03 (band 5e-02 declared),bf16
 swa_dit_softmax_bwd_sm90_cuda,cuda,swa_dit,miniworld_engine/kernels/swa_dit/cuda/swa_attn.cu,ok,rel dq=5.60e-03 dmod=3.64e-03 dw_qkv=5.72e-03 (band 5e-02 declared),bf16
-swa_dit_output_bwd_sm90_cuda,cuda,swa_dit,miniworld_engine/kernels/swa_dit/cuda/swa_slab.cu,ok,rel dq=6.18e-03 dmod=3.40e-03 dw_out=4.66e-03 dw_gate=4.68e-03 (band 5e-02 declared),bf16
-swa_dit_inproj_bwd_sm90_cuda,cuda,swa_dit,miniworld_engine/kernels/swa_dit/cuda/swa_slab.cu,ok,rel dq=6.18e-03 dmod=3.40e-03 dw_qkv=5.72e-03 dw_gate=4.68e-03 (band 5e-02 declared),bf16
+swa_dit_output_bwd_sm90_cuda,cuda,swa_dit,miniworld_engine/kernels/swa_dit/cuda/swa_slab.cu,ok,rel dq=6.18e-03 dmod=3.40e-03 dw_out=4.32e-03 dw_gate=4.86e-03 (band 5e-02 declared),bf16
+swa_dit_inproj_bwd_sm90_cuda,cuda,swa_dit,miniworld_engine/kernels/swa_dit/cuda/swa_slab.cu,ok,rel dq=6.18e-03 dmod=3.40e-03 dw_qkv=6.74e-03 dw_gate=4.86e-03 (band 5e-02 declared),bf16
 swa_dit_inproj_fwd_rows_sm90_cuda,cuda,swa_dit,miniworld_engine/kernels/swa_dit/cuda/swa_slab.cu,ok,rel out=5.90e-03 (band 5e-02 declared),bf16
 swa_dit_output_fwd_rows_sm90_cuda,cuda,swa_dit,miniworld_engine/kernels/swa_dit/cuda/swa_slab.cu,ok,rel out=5.90e-03 (band 5e-02 declared),bf16
 swa_dit_swiglu_fwd_rows_sm90_cuda,cuda,swa_dit,miniworld_engine/kernels/swa_dit/cuda/swa_slab.cu,ok,rel out=5.90e-03 (band 5e-02 declared),bf16

{miniworld_engine-2.3.10 → miniworld_engine-2.3.12}/src/miniworld_engine/autotune/records.py

@@ -170,7 +170,7 @@ def missing(op: str, name: str, capability: tuple[int, int], dtype: str = "bf16"
     table does not know has no kernels to vouch for: its only entry is a note naming the op."""
     kernels = triton_kernels(op, dtype, mode)
     if not kernels:
-        return (f"<no Triton kernel of {op!r} declares {dtype}>",)
+        return (f"<no portable kernel of {op!r} declares {dtype}>",)
     rows = _load(card_key(name, capability))
     return tuple(k for k in kernels if rows.get((k, dtype)) != "ok")
 

{miniworld_engine-2.3.10 → miniworld_engine-2.3.12}/src/miniworld_engine/build/bundle_coverage.py

@@ -147,6 +147,19 @@ def cases() -> list[dict]:
     # The Anthropic OPM path (integrations/anthropic_msa) finishes with THIS repo's csrc/opm_epilogue.cu, built once under the name
     # ``miniworld_opm_epilogue`` (no variant suffix: the ``_native*`` ones belong to integrations/opm_train). Needs no payload to build.
     out.append({"kind": "anthropic_ext", "train": False, "dtype": "bfloat16", "heavy": False})
+    # La-Proteina (kernels/pair_embed_geom, pair_embed_adaln; the token DiT block at its widths: 768 / 12 heads x 64 / pair 256 / cond 256 or 128 /
+    # transition x4): the pair builders once, and the batched-pairs block (block_psb under serves_batched) inference and training at the padded
+    # token lengths, B = 5 at 384 (the TOKEN_EMPTY_UNIFORM build) and a larger B elsewhere. H100 only: the others report a skip.
+    out.append({"kind": "la_proteina_pair", "train": True, "dtype": "bfloat16", "heavy": False})
+    for train in (False, True):
+        for length in (128, 256, 384, 512):
+            for d_cond in (256, 128):
+                for qk in (True, False):
+                    if not qk and d_cond != 256:
+                        continue
+                    for b in ((5, 2) if length == 384 else (2, 12)):
+                        out.append({"kind": "la_proteina_dit", "L": length, "train": train, "dtype": "bfloat16", "d_cond": d_cond,  # noqa: PERF401
+                                    "qk": qk, "batch": b, "heavy": False})
     for i, c in enumerate(out):
         c["id"] = i
     return out
@@ -320,6 +333,12 @@ def _run_block(c: dict) -> None:
     elif c["kind"] == "model_variants":
         _model_variants()
         return
+    elif c["kind"] == "la_proteina_pair":
+        _la_proteina_pair()
+        return
+    elif c["kind"] == "la_proteina_dit":
+        _la_proteina_dit(c)
+        return
     elif c["kind"] == "sampler_ops":
         _sampler_ops()
         return
@@ -346,6 +365,65 @@ def _run_block(c: dict) -> None:
     _step(model, args, c["train"], autocast)
 
 
+def _la_proteina_pair() -> None:
+    """La-Proteina's pair builders, every extension variant once: the encoder pair (inference; training with and without the folded LayerNorm), the
+    decoder pair (training, with and without it) with the table-gradient backward, and the initial pair builder (self-conditioning on / off)."""
+    import torch
+
+    from miniworld_engine.kernels.pair_embed_adaln import pair_embed_adaln
+    from miniworld_engine.kernels.pair_embed_geom import pair_embed_ca, pair_embed_geom, pair_embed_geom_train
+    from miniworld_engine.kernels.pair_embed_geom.cuda import sm90 as geom
+
+    if torch.cuda.get_device_capability() != (9, 0):
+        raise RuntimeError("La-Proteina's pair builders are H100 (sm_90a) kernels")
+    dev = "cuda"
+    for b, n in ((48, 128), (2, 512)):
+        bb = torch.randn(b, n, 4, 3, device=dev) * 0.5
+        flags = [torch.ones(b, n, device=dev, dtype=torch.bool) for _ in range(3)]
+        idx, chains = torch.arange(1, n + 1, device=dev).expand(b, n), torch.zeros(b, n, device=dev, dtype=torch.int32)
+        w = torch.randn(256, 317, device=dev, requires_grad=True)
+        wc = torch.randn(256, 157, device=dev, requires_grad=True)
+        ca = bb[:, :, 1].contiguous()
+        with torch.no_grad():
+            pair_embed_geom(w.detach(), bb, flags[0], flags[1], flags[2], idx, chains)
+        for eps in (None, 1e-5):
+            pair_embed_geom_train(w, bb, flags[0], flags[1], flags[2], idx, chains, ln_eps=eps).float().sum().backward()
+            pair_embed_ca(wc, ca, flags[2], ln_eps=eps).float().sum().backward()
+        for impl in (0, 1):   # the shared-memory read-modify-write backward is the A/B switch MINIWORLD_PAIR_EMBED_BWD_IMPL=0
+            pair, codes = geom.forward(w.detach().t().to(torch.bfloat16).contiguous(), bb, flags[0].to(torch.uint8), flags[1].to(torch.uint8),
+                                       flags[2].to(torch.uint8), idx.to(torch.int32), chains, torch.linspace(0.1, 2, 20, device=dev),
+                                       torch.linspace(-torch.pi, torch.pi, 20, device=dev), with_codes=True)
+            geom.backward(pair, codes, 0, impl=impl)
+        ln_w, ln_b = torch.ones(256, device=dev, requires_grad=True), torch.zeros(256, device=dev, requires_grad=True)
+        wa = torch.randn(256, 187, device=dev, requires_grad=True)
+        gamma, beta = torch.rand(b, 256, device=dev, requires_grad=True), torch.randn(b, 256, device=dev, requires_grad=True)
+        for sc in (None, ca):
+            pair_embed_adaln(wa, ln_w, ln_b, gamma, beta, ca, sc, idx, flags[2]).float().sum().backward()
+    torch.cuda.synchronize()
+
+
+def _la_proteina_dit(c: dict) -> None:
+    """The token DiT block at La-Proteina's widths over ``batch`` independent proteins, each with its own pair (masked), bf16."""
+    import torch
+
+    from miniworld_engine.modules.dit import DiTBlock
+    from miniworld_engine.modules.exceptions import ImplementationType as MI
+
+    b, n, d_cond, train = c["batch"], c["L"], c["d_cond"], c["train"]
+    model = DiTBlock(d_single=768, d_cond=d_cond, d_pair=256, n_head=12, n=4, use_qk_norm=c["qk"], implementation=MI.MINIWORLD).cuda().bfloat16()
+    _randomize(model)
+    model.train(train)
+    single = torch.randn(1, b, n, 768, device="cuda", dtype=torch.bfloat16, requires_grad=train)
+    cond = torch.randn(1, b, n, d_cond, device="cuda", dtype=torch.bfloat16, requires_grad=train)
+    pair = torch.randn(b, n, n, 256, device="cuda", dtype=torch.bfloat16, requires_grad=train)
+    mask = torch.arange(n, device="cuda")[None] < torch.randint(n // 2, n + 1, (b, 1), device="cuda")
+    with torch.set_grad_enabled(train):
+        out = model(single, cond, pair, mask)
+        if train:
+            out.float().sum().backward()
+    torch.cuda.synchronize()
+
+
 def _swa_stages() -> None:
     """Load every stage extension of the fused SWA atom block once (a stage that fails to build is an error, not a silent fallback)."""
     from miniworld_engine.kernels.swa_dit.cuda import loader
@@ -459,7 +537,7 @@ def serve() -> int:
     """Worker loop: one JSON descriptor per stdin line, one ``RESULT <json>`` line per descriptor on stdout."""
     from miniworld_engine import settings
 
-    settings.configure(triton_on_hopper_blackwell=False, force_triton_kernels=False)
+    settings.configure(portable_on_hopper_blackwell=False, force_portable_kernels=False)
     for line in sys.stdin:
         line = line.strip()
         if not line:

{miniworld_engine-2.3.10 → miniworld_engine-2.3.12}/src/miniworld_engine/cli.py

@@ -192,7 +192,7 @@ PINS: dict[str, tuple[tuple, tuple[str, ...], tuple[str, ...]]] = {
     #: would just rebuild the same shard under a second name.
     "dropout": ((0.25,), ("triangle_multiplication",), ("training",)),
     # transition's forward runs the hand-CUDA fused b2b when it applies, and that kernel is not
-    # autotuned -- so the TRITON expand-gate forward never fires during a build and
+    # autotuned -- so the PORTABLE expand-gate forward never fires during a build and
     # transition_expand_gate_fwd stays absent from the cache, while its backward (which has no CUDA
     # counterpart) is captured normally. The CUDA path only covers fixed shapes, so production
     # falls back to the triton one elsewhere and finds nothing cached. Sweeping it off captures
@@ -837,10 +837,7 @@ def _reject_unknown_build_target(args: argparse.Namespace, repo: Path) -> int:
 
 
 def _kernels_choice(args: argparse.Namespace) -> str:
-    """``build --kernels {auto,native,portable}``; the hidden ``--backend {auto,native,triton}`` is its deprecated spelling (``triton`` is ``portable``)."""
-    old = getattr(args, "backend", None)
-    if old is not None:
-        return {"triton": "portable"}.get(old, old)
+    """``build --kernels {auto,native,portable}``."""
     return getattr(args, "kernels", None) or "auto"
 
 
@@ -899,7 +896,7 @@ def cmd_build(args: argparse.Namespace) -> int:
         from miniworld_engine.autotune import policy
         units = policy.filter_op_units(units)
         if not units:
-            print(f"no triton op with a driver matched {args.case!r}", file=sys.stderr)
+            print(f"no kernel op with a driver matched {args.case!r}", file=sys.stderr)
             return None
         if only and len(only) > len({u.op for u in units}):
             # Named and not built is not the same as not named. A kernel with no driver, or none
@@ -923,7 +920,7 @@ def cmd_build(args: argparse.Namespace) -> int:
 
     # A NAMED target drives its MODULE: the two name spaces are different -- `build <case>` names a
     # module and `--per-op <kernel>` names a kernel -- so running the op sweep for a case name
-    # filters `op_units` by a name no kernel has and returns "no triton op with a driver matched".
+    # filters `op_units` by a name no kernel has and returns "no kernel op with a driver matched".
     # Which is what this did for one commit, turning `build gated_projection grid` into exit 2.
     # The MODULE pass is the default now, including for `all`. It used to be the op pass, because
     # the module pass reached only the 48 of 91 kernels some module happened to dispatch to while
@@ -1751,7 +1748,7 @@ def build_parser() -> argparse.ArgumentParser:
     cap.add_argument("--shards", default="~/.cache/miniworld-shards", help="where shards go")
     cap.add_argument("--impl", default="miniworld",
                      help="implementations=[...]. Use miniworld: for triangle_multiplication, "
-                          "'triton' builds the fp32 dtv1 baseline and captures nothing")
+                          "'portable' builds the fp32 baseline and captures nothing")
     cap.add_argument("--no-sweep-dispatch", action="store_true",
                      help="capture only the branch this card picks (leaves the other uncached)")
     cap.add_argument("--resume", action="store_true", help="skip shards that already exist")
@@ -1801,9 +1798,8 @@ def build_parser() -> argparse.ArgumentParser:
     bld.add_argument("config_type", nargs="?", default=DEFAULT_CONFIG_SET,
                      help="config set: a directory of <op>.csv files, or a short name resolving to configs/<name> (e.g. accuracy). Every kernel's grid comes from here.")
     bld.add_argument("--kernels", choices=("auto", "native", "portable"), default="auto",
-                     help="auto follows module dispatch; native tunes the hand-written CUDA per-op drivers, portable the Triton ones "
+                     help="auto follows module dispatch; native tunes the hand-written CUDA per-op drivers, portable the portable-kernel ones "
                           "(the portable family is internal: it is what miniworld routes to on a card with a record)")
-    bld.add_argument("--backend", choices=("auto", "native", "triton"), default=None, help=argparse.SUPPRESS)  # deprecated spelling of --kernels
     bld.add_argument("--include-alternatives", action="store_true",
                      help="also tune forced fallback backends; default follows production dispatch")
     bld.add_argument("--mode", choices=("both", "train", "eval"), default="both",

{miniworld_engine-2.3.10 → miniworld_engine-2.3.12}/src/miniworld_engine/integrations/anthropic_msa.py

@@ -123,7 +123,7 @@ def wanted(implementation) -> bool:
         return True
     if implementation != ImplementationType.MINIWORLD:
         return False
-    return bool(payload_dir()) and (not settings.current().force_triton_kernels)
+    return bool(payload_dir()) and (not settings.current().force_portable_kernels)
 
 
 def _common_refusal(x: torch.Tensor, *, grad: bool) -> str | None:

{miniworld_engine-2.3.10 → miniworld_engine-2.3.12}/src/miniworld_engine/integrations/anthropic_trimul.py

@@ -105,7 +105,7 @@ def wanted(implementation) -> bool:
         return True
     if implementation != ImplementationType.MINIWORLD:
         return False
-    return bool(payload_dir()) and (not settings.current().force_triton_kernels)
+    return bool(payload_dir()) and (not settings.current().force_portable_kernels)
 
 
 def refusal(pair: torch.Tensor, c_z: int, c_hidden: int, *, grad: bool, dropout: bool) -> str | None:

{miniworld_engine-2.3.10 → miniworld_engine-2.3.12}/src/miniworld_engine/integrations/attention_pair_bias_b200.py

@@ -68,7 +68,7 @@ def _eligible(
 ) -> bool:
     if (
         module.implementation != ImplementationType.MINIWORLD
-        or settings.current().force_triton_kernels
+        or settings.current().force_portable_kernels
     ):
         return False
     H, D = module.n_head, single.shape[-1]

{miniworld_engine-2.3.10 → miniworld_engine-2.3.12}/src/miniworld_engine/integrations/bias_only_dit.py

@@ -15,7 +15,7 @@ settings.bias_only_dit_inf3=0 (read per call) keeps the 12-launch cuBLAS + rows
 ``bo_tail2_bf16``) behind bf16 tables (``runner._tables3b``); settings.bias_only_dit_inf3_bf16=0 (read per call) keeps the 12-launch
 bf16 step, which is also what a failed build falls back to.
 
-``serves()`` is the whole gate: no autograd, the engine's kernels (implementation MINIWORLD or TRITON), B200 or H100, bf16 or fp32, the
+``serves()`` is the whole gate: no autograd, the engine's kernels (implementation MINIWORLD), B200 or H100, bf16 or fp32, the
 token widths (768; the attention as 16 heads x 48, 24 x 32, 12 x 64 or 16 x 64 / cond 384 / pair 128 / transition n = 2), B == 1, L a multiple of 128 (up to 768; 3072 on H100: ``core_supported``), a key mask [1, L] or
 none, LayerNorm eps 1e-5. The conditioning may be one per sample or one shared by the samples (sample axis 1 or stride 0).
 settings.bias_only_dit=0 turns it off. Anything else keeps the module's PyTorch composition.
@@ -77,7 +77,7 @@ def serves(module, single, cond, pair, mask=None, bias=None) -> bool:
         return False
     if module.implementation != ImplementationType.MINIWORLD:
         return False
-    if settings.current().force_triton_kernels:
+    if settings.current().force_portable_kernels:
         return False
     dt = single.dtype
     src = (

{miniworld_engine-2.3.10 → miniworld_engine-2.3.12}/src/miniworld_engine/integrations/bias_only_dit_hoist.py

@@ -61,7 +61,7 @@ def serves(implementation, attns, pair, per_sample=False) -> bool:
         return False
     if implementation != ImplementationType.MINIWORLD:
         return False
-    if settings.current().force_triton_kernels:
+    if settings.current().force_portable_kernels:
         return False
     if not pair.is_cuda or torch.cuda.get_device_capability(pair.device) not in ((10, 0), (9, 0)):
         return False

{miniworld_engine-2.3.10 → miniworld_engine-2.3.12}/src/miniworld_engine/integrations/bias_only_dit_train.py

@@ -34,7 +34,7 @@ the dxa GEMM / adaln_a_bwd / dchat + dcg GEMMs / cond_bwd (``bo_bwd_mid`` x 3 wi
 weight gradients and the unfold, ``pair_bias_bwd_fin`` for pair_bias_bwd + finalize (_BWD_PBFIN). settings.bias_only_dit_bwd_fused=0,
 read per call, or a failed build: the per-step launches.
 
-``serves()`` is the whole gate: autograd on, the engine's kernels (implementation MINIWORLD or TRITON), B200, bf16 or fp32 inputs, the
+``serves()`` is the whole gate: autograd on, the engine's kernels (implementation MINIWORLD), B200, bf16 or fp32 inputs, the
 token widths (768; the attention as 16 heads x 48, 24 x 32, 12 x 64 or 16 x 64 / cond 384 / pair 128 / transition n = 2), B == 1, L a multiple of 128 up to 768, a per-sample
 conditioning [A, 1, L, 384], a key mask [1, L] or none, LayerNorm eps 1e-5. settings.bias_only_dit_train=0 turns it off.
 
@@ -105,7 +105,7 @@ def serves(module, single, cond, pair, mask=None, bias=None, per_sample=False) -
         return False
     if module.implementation != ImplementationType.MINIWORLD:
         return False
-    if settings.current().force_triton_kernels:
+    if settings.current().force_portable_kernels:
         return False
     from miniworld_engine.kernels.bias_only_dit.cuda import h100_enabled
 

{miniworld_engine-2.3.10 → miniworld_engine-2.3.12}/src/miniworld_engine/integrations/local_dit.py

@@ -18,7 +18,7 @@ depend on what the attention attends to. The attention and the pair bias are ``s
 the trunked pair ``[nwin, 32, 128, 16]`` (bias ``LN(z) Wb`` per window, shared by the samples), no N x N tensor anywhere.
 Both residuals are inside post_fwd (the block returns the stream).
 
-``serves()`` is the whole gate: B200, the engine's kernel backend (implementation MINIWORLD or TRITON), bf16 single / cond / pair,
+``serves()`` is the whole gate: B200, the engine's kernel backend (implementation MINIWORLD), bf16 single / cond / pair,
 the atom widths (d_single = d_cond = 128, 4 heads x 32, d_pair = 16, transition n = 2, no QK-norm), B == 1, any N, a [B, N] bool key
 mask or none, LayerNorm eps 1e-5. N is padded to a multiple of 128 for the row kernels (zero single / cond rows, the padded atoms
 masked as keys, the pair padded with zero windows) and the first N rows come back. ``settings.local_dit_sm100=0`` turns it off. A
@@ -358,13 +358,13 @@ def serves(module, single, cond, pair, mask, rope=None) -> bool:
     if (
         (_FAILED32 if fp32 else _FAILED)
         or (not settings.current().local_dit_sm100)
-        or settings.current().force_triton_kernels
+        or settings.current().force_portable_kernels
     ):
         return False
     from miniworld_engine.modules.dispatch import KernelBackend
 
     a, tr = module.attention, module.transition
-    if a._backend != KernelBackend.TRITON or a.use_qk_norm:
+    if a._backend != KernelBackend.PORTABLE or a.use_qk_norm:
         return False
     cap = torch.cuda.get_device_capability(single.device) if single.is_cuda else None
     if cap != (10, 0):
@@ -449,7 +449,7 @@ def _require_hopper_training(module, single, cond, pair, mask):
 
     a, tr = module.attention, module.transition
     if (
-        a._backend != KernelBackend.TRITON
+        a._backend != KernelBackend.PORTABLE
         or not torch.is_grad_enabled()
         or not single.is_cuda
         or torch.cuda.get_device_capability(single.device) != (9, 0)
@@ -465,7 +465,7 @@ def _require_hopper_training(module, single, cond, pair, mask):
                 f"H100 local atom DiT training requires {detail}; no composed fallback"
             )
 
-    require(not settings.current().force_triton_kernels, "the CUDA engine backend")
+    require(not settings.current().force_portable_kernels, "the CUDA engine backend")
     require(
         module.cross_attention and not a.use_qk_norm, "cross-attention without QK norm"
     )
@@ -550,7 +550,7 @@ def _require_hopper_inference(module, single, cond, pair, mask):
 
     a, tr = module.attention, module.transition
     if (
-        a._backend != KernelBackend.TRITON
+        a._backend != KernelBackend.PORTABLE
         or torch.is_grad_enabled()
         or not single.is_cuda
     ):
@@ -564,7 +564,7 @@ def _require_hopper_inference(module, single, cond, pair, mask):
                 f"H100 local atom DiT inference requires {detail}; no composed fallback"
             )
 
-    require(not settings.current().force_triton_kernels, "the CUDA engine backend")
+    require(not settings.current().force_portable_kernels, "the CUDA engine backend")
     require(
         module.cross_attention and not a.use_qk_norm, "cross-attention without QK norm"
     )