Promote latest kernel artifacts to main

.gitattributes

@@ -1,35 +1,3 @@
-*.7z filter=lfs diff=lfs merge=lfs -text
-*.arrow filter=lfs diff=lfs merge=lfs -text
-*.bin filter=lfs diff=lfs merge=lfs -text
-*.bz2 filter=lfs diff=lfs merge=lfs -text
-*.ckpt filter=lfs diff=lfs merge=lfs -text
-*.ftz filter=lfs diff=lfs merge=lfs -text
-*.gz filter=lfs diff=lfs merge=lfs -text
-*.h5 filter=lfs diff=lfs merge=lfs -text
-*.joblib filter=lfs diff=lfs merge=lfs -text
-*.lfs.* filter=lfs diff=lfs merge=lfs -text
-*.mlmodel filter=lfs diff=lfs merge=lfs -text
-*.model filter=lfs diff=lfs merge=lfs -text
-*.msgpack filter=lfs diff=lfs merge=lfs -text
-*.npy filter=lfs diff=lfs merge=lfs -text
-*.npz filter=lfs diff=lfs merge=lfs -text
-*.onnx filter=lfs diff=lfs merge=lfs -text
-*.ot filter=lfs diff=lfs merge=lfs -text
-*.parquet filter=lfs diff=lfs merge=lfs -text
-*.pb filter=lfs diff=lfs merge=lfs -text
-*.pickle filter=lfs diff=lfs merge=lfs -text
-*.pkl filter=lfs diff=lfs merge=lfs -text
-*.pt filter=lfs diff=lfs merge=lfs -text
-*.pth filter=lfs diff=lfs merge=lfs -text
-*.rar filter=lfs diff=lfs merge=lfs -text
-*.safetensors filter=lfs diff=lfs merge=lfs -text
-saved_model/**/* filter=lfs diff=lfs merge=lfs -text
-*.tar.* filter=lfs diff=lfs merge=lfs -text
-*.tar filter=lfs diff=lfs merge=lfs -text
-*.tflite filter=lfs diff=lfs merge=lfs -text
-*.tgz filter=lfs diff=lfs merge=lfs -text
-*.wasm filter=lfs diff=lfs merge=lfs -text
-*.xz filter=lfs diff=lfs merge=lfs -text
-*.zip filter=lfs diff=lfs merge=lfs -text
-*.zst filter=lfs diff=lfs merge=lfs -text
-*tfevents* filter=lfs diff=lfs merge=lfs -text
+*.so filter=lfs diff=lfs merge=lfs -text
+*.pyd filter=lfs diff=lfs merge=lfs -text
+*.dylib filter=lfs diff=lfs merge=lfs -text

README.md

@@ -1,9 +0,0 @@
-# flashrt/linear-attention-seq-state
-
-This repository is a compatibility mirror for older `kernels` clients
-that resolve repositories through the default Hugging Face model repo API.
-
-Canonical Kernel Hub repo: https://huggingface.co/kernels/flashrt/linear-attention-seq-state
-
-Do not edit this mirror by hand. It is generated from the Kernel Hub
-`vN` branches and contains the same `build/**` artifacts.

build/torch211-cxx11-cu128-x86_64-linux/__init__.py

@@ -0,0 +1,54 @@
+"""FlashRT sequential state-scan kernels for linear attention."""
+
+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("gated_delta_recurrent_seq_bf16"))
+def _gated_delta_recurrent_seq_fake(
+    q: torch.Tensor,
+    k: torch.Tensor,
+    v: torch.Tensor,
+    g: torch.Tensor,
+    beta: torch.Tensor,
+    state: torch.Tensor,
+    out: torch.Tensor,
+    use_qk_l2norm: bool = False,
+) -> None:
+    if q.dim() != 3 or k.shape != q.shape or v.shape != q.shape:
+        raise RuntimeError("q/k/v must have shape (S,H,D)")
+    if g.shape != (q.shape[0], q.shape[1]) or beta.shape != g.shape:
+        raise RuntimeError("g/beta must have shape (S,H)")
+    if state.shape != (q.shape[1], q.shape[2], q.shape[2]) or out.shape != q.shape:
+        raise RuntimeError("state/out shape mismatch")
+    return None
+
+
+def gated_delta_recurrent_seq_bf16(
+    q: torch.Tensor,
+    k: torch.Tensor,
+    v: torch.Tensor,
+    g: torch.Tensor,
+    beta: torch.Tensor,
+    state: torch.Tensor,
+    *,
+    out: Optional[torch.Tensor] = None,
+    use_qk_l2norm: bool = False,
+) -> tuple[torch.Tensor, torch.Tensor]:
+    """Scan a full `(S,H,128)` Gated DeltaNet sequence in one launch.
+
+    `state` is updated in place with the final `(H,128,128)` state.
+    """
+
+    if out is None:
+        out = torch.empty_like(q)
+    ops.gated_delta_recurrent_seq_bf16(q, k, v, g, beta, state, out, bool(use_qk_l2norm))
+    return out, state
+
+
+__all__ = ["gated_delta_recurrent_seq_bf16"]

build/torch211-cxx11-cu128-x86_64-linux/_linear_attention_seq_state_cuda_dc145d0.abi3.so

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fe0391f1ed2474da4990a40bbfb4bf9c2437135ea9c178f80316b5aefbc79dde
+size 246792

build/torch211-cxx11-cu128-x86_64-linux/_ops.py

@@ -0,0 +1,9 @@
+import torch
+from . import _linear_attention_seq_state_cuda_dc145d0
+ops = torch.ops._linear_attention_seq_state_cuda_dc145d0
+
+def add_op_namespace_prefix(op_name: str):
+    """
+    Prefix op by namespace.
+    """
+    return f"_linear_attention_seq_state_cuda_dc145d0::{op_name}"

build/torch211-cxx11-cu128-x86_64-linux/linear_attention_seq_state/__init__.py

@@ -0,0 +1,26 @@
+import ctypes
+import importlib.util
+import sys
+from pathlib import Path
+from types import ModuleType
+
+
+def _import_from_path(file_path: Path) -> ModuleType:
+    # We cannot use the module name as-is, after adding it to `sys.modules`,
+    # it would also be used for other imports. So, we make a module name that
+    # depends on the path for it to be unique using the hex-encoded hash of
+    # the path.
+    path_hash = "{:x}".format(ctypes.c_size_t(hash(file_path.absolute())).value)
+    module_name = path_hash
+    spec = importlib.util.spec_from_file_location(module_name, file_path)
+    if spec is None:
+        raise ImportError(f"Cannot load spec for {module_name} from {file_path}")
+    module = importlib.util.module_from_spec(spec)
+    if module is None:
+        raise ImportError(f"Cannot load module {module_name} from spec")
+    sys.modules[module_name] = module
+    spec.loader.exec_module(module)  # type: ignore
+    return module
+
+
+globals().update(vars(_import_from_path(Path(__file__).parent.parent / "__init__.py")))

build/torch211-cxx11-cu128-x86_64-linux/metadata.json

@@ -0,0 +1,23 @@
+{
+  "name": "linear-attention-seq-state",
+  "id": "_linear_attention_seq_state_cuda_dc145d0",
+  "version": 1,
+  "license": "Apache-2.0",
+  "python-depends": [],
+  "backend": {
+    "type": "cuda",
+    "archs": [
+      "10.0",
+      "10.1",
+      "12.0+PTX",
+      "7.0",
+      "7.2",
+      "7.5",
+      "8.0",
+      "8.6",
+      "8.7",
+      "8.9",
+      "9.0"
+    ]
+  }
+}

build/torch211-cxx11-cu130-x86_64-linux/__init__.py

@@ -0,0 +1,54 @@
+"""FlashRT sequential state-scan kernels for linear attention."""
+
+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("gated_delta_recurrent_seq_bf16"))
+def _gated_delta_recurrent_seq_fake(
+    q: torch.Tensor,
+    k: torch.Tensor,
+    v: torch.Tensor,
+    g: torch.Tensor,
+    beta: torch.Tensor,
+    state: torch.Tensor,
+    out: torch.Tensor,
+    use_qk_l2norm: bool = False,
+) -> None:
+    if q.dim() != 3 or k.shape != q.shape or v.shape != q.shape:
+        raise RuntimeError("q/k/v must have shape (S,H,D)")
+    if g.shape != (q.shape[0], q.shape[1]) or beta.shape != g.shape:
+        raise RuntimeError("g/beta must have shape (S,H)")
+    if state.shape != (q.shape[1], q.shape[2], q.shape[2]) or out.shape != q.shape:
+        raise RuntimeError("state/out shape mismatch")
+    return None
+
+
+def gated_delta_recurrent_seq_bf16(
+    q: torch.Tensor,
+    k: torch.Tensor,
+    v: torch.Tensor,
+    g: torch.Tensor,
+    beta: torch.Tensor,
+    state: torch.Tensor,
+    *,
+    out: Optional[torch.Tensor] = None,
+    use_qk_l2norm: bool = False,
+) -> tuple[torch.Tensor, torch.Tensor]:
+    """Scan a full `(S,H,128)` Gated DeltaNet sequence in one launch.
+
+    `state` is updated in place with the final `(H,128,128)` state.
+    """
+
+    if out is None:
+        out = torch.empty_like(q)
+    ops.gated_delta_recurrent_seq_bf16(q, k, v, g, beta, state, out, bool(use_qk_l2norm))
+    return out, state
+
+
+__all__ = ["gated_delta_recurrent_seq_bf16"]

build/torch211-cxx11-cu130-x86_64-linux/_linear_attention_seq_state_cuda_dc145d0.abi3.so

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:307c8f19b1b276ffbd5816f4a13752496316825999ef1e4575deda9daf301312
+size 239784

build/torch211-cxx11-cu130-x86_64-linux/_ops.py

@@ -0,0 +1,9 @@
+import torch
+from . import _linear_attention_seq_state_cuda_dc145d0
+ops = torch.ops._linear_attention_seq_state_cuda_dc145d0
+
+def add_op_namespace_prefix(op_name: str):
+    """
+    Prefix op by namespace.
+    """
+    return f"_linear_attention_seq_state_cuda_dc145d0::{op_name}"

build/torch211-cxx11-cu130-x86_64-linux/linear_attention_seq_state/__init__.py

@@ -0,0 +1,26 @@
+import ctypes
+import importlib.util
+import sys
+from pathlib import Path
+from types import ModuleType
+
+
+def _import_from_path(file_path: Path) -> ModuleType:
+    # We cannot use the module name as-is, after adding it to `sys.modules`,
+    # it would also be used for other imports. So, we make a module name that
+    # depends on the path for it to be unique using the hex-encoded hash of
+    # the path.
+    path_hash = "{:x}".format(ctypes.c_size_t(hash(file_path.absolute())).value)
+    module_name = path_hash
+    spec = importlib.util.spec_from_file_location(module_name, file_path)
+    if spec is None:
+        raise ImportError(f"Cannot load spec for {module_name} from {file_path}")
+    module = importlib.util.module_from_spec(spec)
+    if module is None:
+        raise ImportError(f"Cannot load module {module_name} from spec")
+    sys.modules[module_name] = module
+    spec.loader.exec_module(module)  # type: ignore
+    return module
+
+
+globals().update(vars(_import_from_path(Path(__file__).parent.parent / "__init__.py")))

build/torch211-cxx11-cu130-x86_64-linux/metadata.json

@@ -0,0 +1,22 @@
+{
+  "name": "linear-attention-seq-state",
+  "id": "_linear_attention_seq_state_cuda_dc145d0",
+  "version": 1,
+  "license": "Apache-2.0",
+  "python-depends": [],
+  "backend": {
+    "type": "cuda",
+    "archs": [
+      "10.0",
+      "11.0",
+      "12.0",
+      "12.1+PTX",
+      "7.5",
+      "8.0",
+      "8.6",
+      "8.7",
+      "8.9",
+      "9.0"
+    ]
+  }
+}

build/torch212-cxx11-cu130-x86_64-linux/__init__.py

@@ -0,0 +1,54 @@
+"""FlashRT sequential state-scan kernels for linear attention."""
+
+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("gated_delta_recurrent_seq_bf16"))
+def _gated_delta_recurrent_seq_fake(
+    q: torch.Tensor,
+    k: torch.Tensor,
+    v: torch.Tensor,
+    g: torch.Tensor,
+    beta: torch.Tensor,
+    state: torch.Tensor,
+    out: torch.Tensor,
+    use_qk_l2norm: bool = False,
+) -> None:
+    if q.dim() != 3 or k.shape != q.shape or v.shape != q.shape:
+        raise RuntimeError("q/k/v must have shape (S,H,D)")
+    if g.shape != (q.shape[0], q.shape[1]) or beta.shape != g.shape:
+        raise RuntimeError("g/beta must have shape (S,H)")
+    if state.shape != (q.shape[1], q.shape[2], q.shape[2]) or out.shape != q.shape:
+        raise RuntimeError("state/out shape mismatch")
+    return None
+
+
+def gated_delta_recurrent_seq_bf16(
+    q: torch.Tensor,
+    k: torch.Tensor,
+    v: torch.Tensor,
+    g: torch.Tensor,
+    beta: torch.Tensor,
+    state: torch.Tensor,
+    *,
+    out: Optional[torch.Tensor] = None,
+    use_qk_l2norm: bool = False,
+) -> tuple[torch.Tensor, torch.Tensor]:
+    """Scan a full `(S,H,128)` Gated DeltaNet sequence in one launch.
+
+    `state` is updated in place with the final `(H,128,128)` state.
+    """
+
+    if out is None:
+        out = torch.empty_like(q)
+    ops.gated_delta_recurrent_seq_bf16(q, k, v, g, beta, state, out, bool(use_qk_l2norm))
+    return out, state
+
+
+__all__ = ["gated_delta_recurrent_seq_bf16"]

build/torch212-cxx11-cu130-x86_64-linux/_linear_attention_seq_state_cuda_dc145d0.abi3.so

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:35325c9c634607ddf35e0cdb9da57ba374aab7990261c190106145fcf42e494a
+size 250696

build/torch212-cxx11-cu130-x86_64-linux/_ops.py

@@ -0,0 +1,9 @@
+import torch
+from . import _linear_attention_seq_state_cuda_dc145d0
+ops = torch.ops._linear_attention_seq_state_cuda_dc145d0
+
+def add_op_namespace_prefix(op_name: str):
+    """
+    Prefix op by namespace.
+    """
+    return f"_linear_attention_seq_state_cuda_dc145d0::{op_name}"

build/torch212-cxx11-cu130-x86_64-linux/linear_attention_seq_state/__init__.py

@@ -0,0 +1,26 @@
+import ctypes
+import importlib.util
+import sys
+from pathlib import Path
+from types import ModuleType
+
+
+def _import_from_path(file_path: Path) -> ModuleType:
+    # We cannot use the module name as-is, after adding it to `sys.modules`,
+    # it would also be used for other imports. So, we make a module name that
+    # depends on the path for it to be unique using the hex-encoded hash of
+    # the path.
+    path_hash = "{:x}".format(ctypes.c_size_t(hash(file_path.absolute())).value)
+    module_name = path_hash
+    spec = importlib.util.spec_from_file_location(module_name, file_path)
+    if spec is None:
+        raise ImportError(f"Cannot load spec for {module_name} from {file_path}")
+    module = importlib.util.module_from_spec(spec)
+    if module is None:
+        raise ImportError(f"Cannot load module {module_name} from spec")
+    sys.modules[module_name] = module
+    spec.loader.exec_module(module)  # type: ignore
+    return module
+
+
+globals().update(vars(_import_from_path(Path(__file__).parent.parent / "__init__.py")))

build/torch212-cxx11-cu130-x86_64-linux/metadata.json

@@ -0,0 +1,22 @@
+{
+  "name": "linear-attention-seq-state",
+  "id": "_linear_attention_seq_state_cuda_dc145d0",
+  "version": 1,
+  "license": "Apache-2.0",
+  "python-depends": [],
+  "backend": {
+    "type": "cuda",
+    "archs": [
+      "10.0",
+      "11.0",
+      "12.0",
+      "12.1+PTX",
+      "7.5",
+      "8.0",
+      "8.6",
+      "8.7",
+      "8.9",
+      "9.0"
+    ]
+  }
+}

build/torch212-cxx11-cu132-x86_64-linux/__init__.py

@@ -0,0 +1,54 @@
+"""FlashRT sequential state-scan kernels for linear attention."""
+
+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("gated_delta_recurrent_seq_bf16"))
+def _gated_delta_recurrent_seq_fake(
+    q: torch.Tensor,
+    k: torch.Tensor,
+    v: torch.Tensor,
+    g: torch.Tensor,
+    beta: torch.Tensor,
+    state: torch.Tensor,
+    out: torch.Tensor,
+    use_qk_l2norm: bool = False,
+) -> None:
+    if q.dim() != 3 or k.shape != q.shape or v.shape != q.shape:
+        raise RuntimeError("q/k/v must have shape (S,H,D)")
+    if g.shape != (q.shape[0], q.shape[1]) or beta.shape != g.shape:
+        raise RuntimeError("g/beta must have shape (S,H)")
+    if state.shape != (q.shape[1], q.shape[2], q.shape[2]) or out.shape != q.shape:
+        raise RuntimeError("state/out shape mismatch")
+    return None
+
+
+def gated_delta_recurrent_seq_bf16(
+    q: torch.Tensor,
+    k: torch.Tensor,
+    v: torch.Tensor,
+    g: torch.Tensor,
+    beta: torch.Tensor,
+    state: torch.Tensor,
+    *,
+    out: Optional[torch.Tensor] = None,
+    use_qk_l2norm: bool = False,
+) -> tuple[torch.Tensor, torch.Tensor]:
+    """Scan a full `(S,H,128)` Gated DeltaNet sequence in one launch.
+
+    `state` is updated in place with the final `(H,128,128)` state.
+    """
+
+    if out is None:
+        out = torch.empty_like(q)
+    ops.gated_delta_recurrent_seq_bf16(q, k, v, g, beta, state, out, bool(use_qk_l2norm))
+    return out, state
+
+
+__all__ = ["gated_delta_recurrent_seq_bf16"]

build/torch212-cxx11-cu132-x86_64-linux/_linear_attention_seq_state_cuda_dc145d0.abi3.so

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b15ccae32f404cded7121d50917afb3697c717d5ca9b1a9335ebbd991dd7dcca
+size 250696

build/torch212-cxx11-cu132-x86_64-linux/_ops.py

@@ -0,0 +1,9 @@
+import torch
+from . import _linear_attention_seq_state_cuda_dc145d0
+ops = torch.ops._linear_attention_seq_state_cuda_dc145d0
+
+def add_op_namespace_prefix(op_name: str):
+    """
+    Prefix op by namespace.
+    """
+    return f"_linear_attention_seq_state_cuda_dc145d0::{op_name}"

build/torch212-cxx11-cu132-x86_64-linux/linear_attention_seq_state/__init__.py

@@ -0,0 +1,26 @@
+import ctypes
+import importlib.util
+import sys
+from pathlib import Path
+from types import ModuleType
+
+
+def _import_from_path(file_path: Path) -> ModuleType:
+    # We cannot use the module name as-is, after adding it to `sys.modules`,
+    # it would also be used for other imports. So, we make a module name that
+    # depends on the path for it to be unique using the hex-encoded hash of
+    # the path.
+    path_hash = "{:x}".format(ctypes.c_size_t(hash(file_path.absolute())).value)
+    module_name = path_hash
+    spec = importlib.util.spec_from_file_location(module_name, file_path)
+    if spec is None:
+        raise ImportError(f"Cannot load spec for {module_name} from {file_path}")
+    module = importlib.util.module_from_spec(spec)
+    if module is None:
+        raise ImportError(f"Cannot load module {module_name} from spec")
+    sys.modules[module_name] = module
+    spec.loader.exec_module(module)  # type: ignore
+    return module
+
+
+globals().update(vars(_import_from_path(Path(__file__).parent.parent / "__init__.py")))

build/torch212-cxx11-cu132-x86_64-linux/metadata.json

@@ -0,0 +1,22 @@
+{
+  "name": "linear-attention-seq-state",
+  "id": "_linear_attention_seq_state_cuda_dc145d0",
+  "version": 1,
+  "license": "Apache-2.0",
+  "python-depends": [],
+  "backend": {
+    "type": "cuda",
+    "archs": [
+      "10.0",
+      "11.0",
+      "12.0",
+      "12.1+PTX",
+      "7.5",
+      "8.0",
+      "8.6",
+      "8.7",
+      "8.9",
+      "9.0"
+    ]
+  }
+}