Add torch213-cxx11-cu130-aarch64-linux artifact

Browse files

Files changed (5) hide show

build/torch213-cxx11-cu130-aarch64-linux/__init__.py +71 -0
build/torch213-cxx11-cu130-aarch64-linux/_ops.py +6 -0
build/torch213-cxx11-cu130-aarch64-linux/bf16_linear_gemv/__init__.py +14 -0
build/torch213-cxx11-cu130-aarch64-linux/bf16_linear_gemv_source_test.abi3.so +3 -0
build/torch213-cxx11-cu130-aarch64-linux/metadata.json +32 -0

build/torch213-cxx11-cu130-aarch64-linux/__init__.py ADDED Viewed

	@@ -0,0 +1,71 @@

+"""FlashRT BF16 decode GEMV kernels."""
+from __future__ import annotations
+from typing import Optional
+import torch
+from ._ops import add_op_namespace_prefix, ops
+@torch.library.register_fake(add_op_namespace_prefix("bf16_decode_gemv_bf16"))
+def _bf16_decode_gemv_fake(
+    x: torch.Tensor,
+    weight: torch.Tensor,
+    alpha: float,
+    variant: int,
+    out: torch.Tensor,
+) -> None:
+    k = x.shape[0] if x.dim() == 1 else x.shape[1]
+    if weight.dim() != 2 or weight.shape[1] != k or out.shape != (weight.shape[0],):
+        raise RuntimeError("expected x (K,) or (1,K), weight (N,K), out (N,)")
+    return None
+@torch.library.register_fake(add_op_namespace_prefix("bf16_decode_gemv_unrolled_bf16"))
+def _bf16_decode_gemv_unrolled_fake(
+    x: torch.Tensor,
+    weight: torch.Tensor,
+    out: torch.Tensor,
+) -> None:
+    k = x.shape[0] if x.dim() == 1 else x.shape[1]
+    if weight.dim() != 2 or weight.shape[1] != k or out.shape != (weight.shape[0],):
+        raise RuntimeError("expected x (K,) or (1,K), weight (N,K), out (N,)")
+    return None
+def bf16_decode_gemv_bf16(
+    x: torch.Tensor,
+    weight: torch.Tensor,
+    *,
+    alpha: float = 1.0,
+    variant: int = 0,
+    out: Optional[torch.Tensor] = None,
+) -> torch.Tensor:
+    """Compute BF16 M=1 decode GEMV: ``out = x @ weight.T * alpha``."""
+    if out is None:
+        out = torch.empty((weight.shape[0],), device=x.device, dtype=torch.bfloat16)
+    ops.bf16_decode_gemv_bf16(x, weight, float(alpha), int(variant), out)
+    return out
+def bf16_decode_gemv_unrolled_bf16(
+    x: torch.Tensor,
+    weight: torch.Tensor,
+    *,
+    out: Optional[torch.Tensor] = None,
+) -> torch.Tensor:
+    """Compute BF16 M=1 GEMV with the unrolled memory-level-parallel kernel."""
+    if out is None:
+        out = torch.empty((weight.shape[0],), device=x.device, dtype=torch.bfloat16)
+    ops.bf16_decode_gemv_unrolled_bf16(x, weight, out)
+    return out
+__all__ = [
+    "bf16_decode_gemv_bf16",
+    "bf16_decode_gemv_unrolled_bf16",
+]

build/torch213-cxx11-cu130-aarch64-linux/_ops.py ADDED Viewed

	@@ -0,0 +1,6 @@

+import torch
+from . import bf16_linear_gemv_source_test
+ops = torch.ops.bf16_linear_gemv_source_test
+def add_op_namespace_prefix(op_name: str):
+    return f"bf16_linear_gemv_source_test::{op_name}"

build/torch213-cxx11-cu130-aarch64-linux/bf16_linear_gemv/__init__.py ADDED Viewed

	@@ -0,0 +1,14 @@

+import ctypes
+import importlib.util
+import sys
+from pathlib import Path
+def _import_from_path(file_path: Path):
+    path_hash = '{:x}'.format(ctypes.c_size_t(hash(file_path.absolute())).value)
+    spec = importlib.util.spec_from_file_location(path_hash, file_path)
+    module = importlib.util.module_from_spec(spec)
+    sys.modules[path_hash] = module
+    spec.loader.exec_module(module)
+    return module
+globals().update(vars(_import_from_path(Path(__file__).parent.parent / '__init__.py')))

build/torch213-cxx11-cu130-aarch64-linux/bf16_linear_gemv_source_test.abi3.so ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:b63706bfe0a497b5c27a6a11b46bcbbe7e7fbfd1d3b608e4c987e45347f48132
+size 242304

build/torch213-cxx11-cu130-aarch64-linux/metadata.json ADDED Viewed

	@@ -0,0 +1,32 @@

+{
+  "name": "bf16-linear-gemv",
+  "id": "bf16_linear_gemv_source_test",
+  "version": 1,
+  "license": "Apache-2.0",
+  "python-depends": [],
+  "backend": {
+    "type": "cuda",
+    "archs": [
+      "11.0"
+    ]
+  },
+  "digest": {
+    "algorithm": "sha256",
+    "files": {
+      "__init__.py": "suVWR4yT6KI3SozcL9Y+/yCd8QFTfoqGQmblO9GCPdc=",
+      "bf16_linear_gemv_source_test.abi3.so": "tjcGv+Ckl7XCemoRtGvLvn5/v9HTtgjkyYfkU0f0gTI=",
+      "_ops.py": "rSHQl/ZTZ94dlDV8sg+HA3shaqmvJGixo+6Rbfj7jvw=",
+      "bf16_linear_gemv/__init__.py": "v6p5XMfQzddhi1fLSAw4HX9CyS0rQsidvu9VsT01xi4="
+    }
+  },
+  "provenance": {
+    "kernel": {
+      "sha": "456d297",
+      "dirty": false
+    },
+    "validation": {
+      "torch": "2.13.0+cu130",
+      "cuda": "13.0"
+    }
+  }
+}