Kernels:

kernels-community
/

layer-norm

Trusted publisher

Kernel card Files Files and versions

xet

Community

kernels-bot commited on 4 days ago

Commit

d2135ff

verified ·

1 Parent(s): a15c370

Uploaded using `kernel-builder`.

Browse files

Files changed (37) hide show

.gitattributes +9 -0
build/torch210-cxx11-cu126-x86_64-linux/{_layer_norm_cuda_86f75d9.abi3.so → _layer_norm_cuda_73ccd0c.abi3.so} +1 -1
build/torch210-cxx11-cu126-x86_64-linux/_ops.py +3 -3
build/torch210-cxx11-cu126-x86_64-linux/metadata.json +1 -1
build/torch210-cxx11-cu128-x86_64-linux/{_layer_norm_cuda_86f75d9.abi3.so → _layer_norm_cuda_73ccd0c.abi3.so} +1 -1
build/torch210-cxx11-cu128-x86_64-linux/_ops.py +3 -3
build/torch210-cxx11-cu128-x86_64-linux/metadata.json +1 -1
build/torch210-cxx11-cu130-x86_64-linux/{_layer_norm_cuda_86f75d9.abi3.so → _layer_norm_cuda_73ccd0c.abi3.so} +1 -1
build/torch210-cxx11-cu130-x86_64-linux/_ops.py +3 -3
build/torch210-cxx11-cu130-x86_64-linux/metadata.json +1 -1
build/torch211-cxx11-cu126-x86_64-linux/{_layer_norm_cuda_86f75d9.abi3.so → _layer_norm_cuda_73ccd0c.abi3.so} +1 -1
build/torch211-cxx11-cu126-x86_64-linux/_ops.py +3 -3
build/torch211-cxx11-cu126-x86_64-linux/metadata.json +1 -1
build/torch211-cxx11-cu128-x86_64-linux/{_layer_norm_cuda_86f75d9.abi3.so → _layer_norm_cuda_73ccd0c.abi3.so} +1 -1
build/torch211-cxx11-cu128-x86_64-linux/_ops.py +3 -3
build/torch211-cxx11-cu128-x86_64-linux/metadata.json +1 -1
build/torch211-cxx11-cu130-x86_64-linux/{_layer_norm_cuda_86f75d9.abi3.so → _layer_norm_cuda_73ccd0c.abi3.so} +1 -1
build/torch211-cxx11-cu130-x86_64-linux/_ops.py +3 -3
build/torch211-cxx11-cu130-x86_64-linux/metadata.json +1 -1
build/torch212-cxx11-cu126-x86_64-linux/__init__.py +26 -0
build/torch212-cxx11-cu126-x86_64-linux/_layer_norm_cuda_73ccd0c.abi3.so +3 -0
build/torch212-cxx11-cu126-x86_64-linux/_ops.py +9 -0
build/torch212-cxx11-cu126-x86_64-linux/layer_norm/__init__.py +26 -0
build/torch212-cxx11-cu126-x86_64-linux/layers.py +51 -0
build/torch212-cxx11-cu126-x86_64-linux/metadata.json +15 -0
build/torch212-cxx11-cu130-x86_64-linux/__init__.py +26 -0
build/torch212-cxx11-cu130-x86_64-linux/_layer_norm_cuda_73ccd0c.abi3.so +3 -0
build/torch212-cxx11-cu130-x86_64-linux/_ops.py +9 -0
build/torch212-cxx11-cu130-x86_64-linux/layer_norm/__init__.py +26 -0
build/torch212-cxx11-cu130-x86_64-linux/layers.py +51 -0
build/torch212-cxx11-cu130-x86_64-linux/metadata.json +17 -0
build/torch212-cxx11-cu132-x86_64-linux/__init__.py +26 -0
build/torch212-cxx11-cu132-x86_64-linux/_layer_norm_cuda_73ccd0c.abi3.so +3 -0
build/torch212-cxx11-cu132-x86_64-linux/_ops.py +9 -0
build/torch212-cxx11-cu132-x86_64-linux/layer_norm/__init__.py +26 -0
build/torch212-cxx11-cu132-x86_64-linux/layers.py +51 -0
build/torch212-cxx11-cu132-x86_64-linux/metadata.json +17 -0

.gitattributes CHANGED Viewed

@@ -96,3 +96,12 @@ build/torch211-cxx11-cu130-aarch64-linux/_layer_norm_cuda_73ccd0c.abi3.so filter
 build/torch212-cxx11-cu126-aarch64-linux/_layer_norm_cuda_73ccd0c.abi3.so filter=lfs diff=lfs merge=lfs -text
 build/torch212-cxx11-cu130-aarch64-linux/_layer_norm_cuda_73ccd0c.abi3.so filter=lfs diff=lfs merge=lfs -text
 build/torch212-cxx11-cu132-aarch64-linux/_layer_norm_cuda_73ccd0c.abi3.so filter=lfs diff=lfs merge=lfs -text

 build/torch212-cxx11-cu126-aarch64-linux/_layer_norm_cuda_73ccd0c.abi3.so filter=lfs diff=lfs merge=lfs -text
 build/torch212-cxx11-cu130-aarch64-linux/_layer_norm_cuda_73ccd0c.abi3.so filter=lfs diff=lfs merge=lfs -text
 build/torch212-cxx11-cu132-aarch64-linux/_layer_norm_cuda_73ccd0c.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch210-cxx11-cu126-x86_64-linux/_layer_norm_cuda_73ccd0c.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch210-cxx11-cu128-x86_64-linux/_layer_norm_cuda_73ccd0c.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch210-cxx11-cu130-x86_64-linux/_layer_norm_cuda_73ccd0c.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch211-cxx11-cu126-x86_64-linux/_layer_norm_cuda_73ccd0c.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch211-cxx11-cu128-x86_64-linux/_layer_norm_cuda_73ccd0c.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch211-cxx11-cu130-x86_64-linux/_layer_norm_cuda_73ccd0c.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch212-cxx11-cu126-x86_64-linux/_layer_norm_cuda_73ccd0c.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch212-cxx11-cu130-x86_64-linux/_layer_norm_cuda_73ccd0c.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch212-cxx11-cu132-x86_64-linux/_layer_norm_cuda_73ccd0c.abi3.so filter=lfs diff=lfs merge=lfs -text

build/torch210-cxx11-cu126-x86_64-linux/{_layer_norm_cuda_86f75d9.abi3.so → _layer_norm_cuda_73ccd0c.abi3.so} RENAMED Viewed

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:956e397d706189303d62d7a3877a4e7a58f551d3611d549bde8889e34fb3a78d
 size 712093824

 version https://git-lfs.github.com/spec/v1
+oid sha256:e1a95a30fff3a4b64414535756ed2d26fc50321c6caeb284a4f1e2e46cfe04dd
 size 712093824

build/torch210-cxx11-cu126-x86_64-linux/_ops.py CHANGED Viewed

@@ -1,9 +1,9 @@
 import torch
-from . import _layer_norm_cuda_86f75d9
-ops = torch.ops._layer_norm_cuda_86f75d9
 def add_op_namespace_prefix(op_name: str):
     """
     Prefix op by namespace.
     """
-    return f"_layer_norm_cuda_86f75d9::{op_name}"

 import torch
+from . import _layer_norm_cuda_73ccd0c
+ops = torch.ops._layer_norm_cuda_73ccd0c
 def add_op_namespace_prefix(op_name: str):
     """
     Prefix op by namespace.
     """
+    return f"_layer_norm_cuda_73ccd0c::{op_name}"

build/torch210-cxx11-cu126-x86_64-linux/metadata.json CHANGED Viewed

@@ -1,6 +1,6 @@
 {
   "name": "layer-norm",
-  "id": "_layer_norm_cuda_86f75d9",
   "version": 1,
   "license": "BSD-3-Clause",
   "python-depends": [],

 {
   "name": "layer-norm",
+  "id": "_layer_norm_cuda_73ccd0c",
   "version": 1,
   "license": "BSD-3-Clause",
   "python-depends": [],

build/torch210-cxx11-cu128-x86_64-linux/{_layer_norm_cuda_86f75d9.abi3.so → _layer_norm_cuda_73ccd0c.abi3.so} RENAMED Viewed

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:aa4fff82fdc3f5566e508d769d5abc40fd49824c1c3362c4f67181f5865c263e
 size 1231419520

 version https://git-lfs.github.com/spec/v1
+oid sha256:f4660dc3f5e3cc4d0531fdebd5d0df2d082b0fad6599a8e63eabcc42b2cedada
 size 1231419520

build/torch210-cxx11-cu128-x86_64-linux/_ops.py CHANGED Viewed

@@ -1,9 +1,9 @@
 import torch
-from . import _layer_norm_cuda_86f75d9
-ops = torch.ops._layer_norm_cuda_86f75d9
 def add_op_namespace_prefix(op_name: str):
     """
     Prefix op by namespace.
     """
-    return f"_layer_norm_cuda_86f75d9::{op_name}"

 import torch
+from . import _layer_norm_cuda_73ccd0c
+ops = torch.ops._layer_norm_cuda_73ccd0c
 def add_op_namespace_prefix(op_name: str):
     """
     Prefix op by namespace.
     """
+    return f"_layer_norm_cuda_73ccd0c::{op_name}"

build/torch210-cxx11-cu128-x86_64-linux/metadata.json CHANGED Viewed

@@ -1,6 +1,6 @@
 {
   "name": "layer-norm",
-  "id": "_layer_norm_cuda_86f75d9",
   "version": 1,
   "license": "BSD-3-Clause",
   "python-depends": [],

 {
   "name": "layer-norm",
+  "id": "_layer_norm_cuda_73ccd0c",
   "version": 1,
   "license": "BSD-3-Clause",
   "python-depends": [],

build/torch210-cxx11-cu130-x86_64-linux/{_layer_norm_cuda_86f75d9.abi3.so → _layer_norm_cuda_73ccd0c.abi3.so} RENAMED Viewed

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a84577823ec575d22141c218eef30298dad1c8b40654c90141a39aa9cbaa1be5
 size 1238402192

 version https://git-lfs.github.com/spec/v1
+oid sha256:61244ac2828b69fe5445df5c6564764eb1b1b80c57312c24836b41595aaf4cc1
 size 1238402192

build/torch210-cxx11-cu130-x86_64-linux/_ops.py CHANGED Viewed

@@ -1,9 +1,9 @@
 import torch
-from . import _layer_norm_cuda_86f75d9
-ops = torch.ops._layer_norm_cuda_86f75d9
 def add_op_namespace_prefix(op_name: str):
     """
     Prefix op by namespace.
     """
-    return f"_layer_norm_cuda_86f75d9::{op_name}"

 import torch
+from . import _layer_norm_cuda_73ccd0c
+ops = torch.ops._layer_norm_cuda_73ccd0c
 def add_op_namespace_prefix(op_name: str):
     """
     Prefix op by namespace.
     """
+    return f"_layer_norm_cuda_73ccd0c::{op_name}"

build/torch210-cxx11-cu130-x86_64-linux/metadata.json CHANGED Viewed

@@ -1,6 +1,6 @@
 {
   "name": "layer-norm",
-  "id": "_layer_norm_cuda_86f75d9",
   "version": 1,
   "license": "BSD-3-Clause",
   "python-depends": [],

 {
   "name": "layer-norm",
+  "id": "_layer_norm_cuda_73ccd0c",
   "version": 1,
   "license": "BSD-3-Clause",
   "python-depends": [],

build/torch211-cxx11-cu126-x86_64-linux/{_layer_norm_cuda_86f75d9.abi3.so → _layer_norm_cuda_73ccd0c.abi3.so} RENAMED Viewed

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:de78905a59142243b204cfda11da6cd85a7df45ddb74e6e297e9f86104c9b675
 size 712082776

 version https://git-lfs.github.com/spec/v1
+oid sha256:a1586a1b500ccc87796c33107b971e38b73d49257aa935a8568c235021490cb9
 size 712082776

build/torch211-cxx11-cu126-x86_64-linux/_ops.py CHANGED Viewed

@@ -1,9 +1,9 @@
 import torch
-from . import _layer_norm_cuda_86f75d9
-ops = torch.ops._layer_norm_cuda_86f75d9
 def add_op_namespace_prefix(op_name: str):
     """
     Prefix op by namespace.
     """
-    return f"_layer_norm_cuda_86f75d9::{op_name}"

 import torch
+from . import _layer_norm_cuda_73ccd0c
+ops = torch.ops._layer_norm_cuda_73ccd0c
 def add_op_namespace_prefix(op_name: str):
     """
     Prefix op by namespace.
     """
+    return f"_layer_norm_cuda_73ccd0c::{op_name}"

build/torch211-cxx11-cu126-x86_64-linux/metadata.json CHANGED Viewed

@@ -1,6 +1,6 @@
 {
   "name": "layer-norm",
-  "id": "_layer_norm_cuda_86f75d9",
   "version": 1,
   "license": "BSD-3-Clause",
   "python-depends": [],

 {
   "name": "layer-norm",
+  "id": "_layer_norm_cuda_73ccd0c",
   "version": 1,
   "license": "BSD-3-Clause",
   "python-depends": [],

build/torch211-cxx11-cu128-x86_64-linux/{_layer_norm_cuda_86f75d9.abi3.so → _layer_norm_cuda_73ccd0c.abi3.so} RENAMED Viewed

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:85e62c45b25b6fd0617b078be0aa55104c1c763ad98ce02dd32db294020c403f
 size 1231408464

 version https://git-lfs.github.com/spec/v1
+oid sha256:be4534f53612397be85ac1dc8075c4dcdb8aba68842de6e909c15ea3ab4cc64d
 size 1231408464

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

@@ -1,9 +1,9 @@
 import torch
-from . import _layer_norm_cuda_86f75d9
-ops = torch.ops._layer_norm_cuda_86f75d9
 def add_op_namespace_prefix(op_name: str):
     """
     Prefix op by namespace.
     """
-    return f"_layer_norm_cuda_86f75d9::{op_name}"

 import torch
+from . import _layer_norm_cuda_73ccd0c
+ops = torch.ops._layer_norm_cuda_73ccd0c
 def add_op_namespace_prefix(op_name: str):
     """
     Prefix op by namespace.
     """
+    return f"_layer_norm_cuda_73ccd0c::{op_name}"

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

@@ -1,6 +1,6 @@
 {
   "name": "layer-norm",
-  "id": "_layer_norm_cuda_86f75d9",
   "version": 1,
   "license": "BSD-3-Clause",
   "python-depends": [],

 {
   "name": "layer-norm",
+  "id": "_layer_norm_cuda_73ccd0c",
   "version": 1,
   "license": "BSD-3-Clause",
   "python-depends": [],

build/torch211-cxx11-cu130-x86_64-linux/{_layer_norm_cuda_86f75d9.abi3.so → _layer_norm_cuda_73ccd0c.abi3.so} RENAMED Viewed

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:63ff31c986d5f8514f0b8cc4a7c336f516a576aa0d2dd1c3f5a0818085035f60
 size 1238395232

 version https://git-lfs.github.com/spec/v1
+oid sha256:9f30404419b58ef0f8bf5176c0d67a176837b70d853f59d7cc326d056e9ab2ee
 size 1238395232

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

@@ -1,9 +1,9 @@
 import torch
-from . import _layer_norm_cuda_86f75d9
-ops = torch.ops._layer_norm_cuda_86f75d9
 def add_op_namespace_prefix(op_name: str):
     """
     Prefix op by namespace.
     """
-    return f"_layer_norm_cuda_86f75d9::{op_name}"

 import torch
+from . import _layer_norm_cuda_73ccd0c
+ops = torch.ops._layer_norm_cuda_73ccd0c
 def add_op_namespace_prefix(op_name: str):
     """
     Prefix op by namespace.
     """
+    return f"_layer_norm_cuda_73ccd0c::{op_name}"

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

@@ -1,6 +1,6 @@
 {
   "name": "layer-norm",
-  "id": "_layer_norm_cuda_86f75d9",
   "version": 1,
   "license": "BSD-3-Clause",
   "python-depends": [],

 {
   "name": "layer-norm",
+  "id": "_layer_norm_cuda_73ccd0c",
   "version": 1,
   "license": "BSD-3-Clause",
   "python-depends": [],

build/torch212-cxx11-cu126-x86_64-linux/__init__.py ADDED Viewed

	@@ -0,0 +1,26 @@

+import torch
+import torch.nn as nn
+from ._ops import ops
+from . import layers
+def dropout_add_ln_fwd(input, gamma, beta, rowscale, colscale, x0_subset, z_subset, dropout_p, epsilon, rowscale_const, z_numrows, gen, residual_in_fp32, is_rms_norm):
+    return ops.dropout_add_ln_fwd(input, gamma, beta, rowscale, colscale, x0_subset, z_subset, dropout_p, epsilon, rowscale_const, z_numrows, gen, residual_in_fp32, is_rms_norm)
+def dropout_add_ln_bwd(dz, dx, x, mu, rsigma, gamma, rowscale, colscale, x0_subset, z_subset, dropout_p, rowscale_const, x0_numrows, has_residual, is_rms_norm):
+    return ops.dropout_add_ln_bwd(dz, dx, x, mu, rsigma, gamma, rowscale, colscale, x0_subset, z_subset, dropout_p, rowscale_const, x0_numrows, has_residual, is_rms_norm)
+def dropout_add_ln_parallel_residual_fwd(input, gamma0, beta0, gamma1, beta1, dropout_p, epsilon, gen, residual_in_fp32, is_rms_norm):
+    return ops.dropout_add_ln_parallel_residual_fwd(input, gamma0, beta0, gamma1, beta1, dropout_p, epsilon, gen, residual_in_fp32, is_rms_norm)
+def dropout_add_ln_parallel_residual_bwd(dz0, dz1, dx, x, mu, rsigma, gamma0, gamma1, dropout_p, has_x1, has_residual, is_rms_norm):
+    return ops.dropout_add_ln_parallel_residual_bwd(dz0, dz1, dx, x, mu, rsigma, gamma0, gamma1, dropout_p, has_x1, has_residual, is_rms_norm)
+__all__ = [
+    "layers",
+    "dropout_add_ln_fwd",
+    "dropout_add_ln_bwd",
+    "dropout_add_ln_parallel_residual_fwd",
+    "dropout_add_ln_parallel_residual_bwd",
+]

build/torch212-cxx11-cu126-x86_64-linux/_layer_norm_cuda_73ccd0c.abi3.so ADDED Viewed

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

+version https://git-lfs.github.com/spec/v1
+oid sha256:c5e4fbafaa6e30159f55100f8fdf7fdcc4a33af09da5345b2327eb01a37b2ad7
+size 712088360

build/torch212-cxx11-cu126-x86_64-linux/_ops.py ADDED Viewed

	@@ -0,0 +1,9 @@

+import torch
+from . import _layer_norm_cuda_73ccd0c
+ops = torch.ops._layer_norm_cuda_73ccd0c
+def add_op_namespace_prefix(op_name: str):
+    """
+    Prefix op by namespace.
+    """
+    return f"_layer_norm_cuda_73ccd0c::{op_name}"

build/torch212-cxx11-cu126-x86_64-linux/layer_norm/__init__.py ADDED Viewed

	@@ -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-cu126-x86_64-linux/layers.py ADDED Viewed

	@@ -0,0 +1,51 @@

+import torch
+import torch.nn as nn
+from ._ops import ops
+class LayerNorm(nn.Module):
+    weight: torch.Tensor
+    variance_epsilon: float
+    def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
+        output = ops.dropout_add_ln_fwd(
+            hidden_states.view(-1, hidden_states.shape[-1]),
+            gamma = self.weight,
+            beta = None,
+            rowscale = None,
+            colscale = None,
+            x0_subset = None,
+            z_subset = None,
+            dropout_p = 0,
+            epsilon = self.variance_epsilon,
+            rowscale_const = 1.0,
+            z_numrows = hidden_states.shape[1],
+            gen = None,
+            residual_in_fp32 = False,
+            is_rms_norm = False,
+        )
+        return output[0].view(hidden_states.shape)
+class LlamaRMSNorm(nn.Module):
+    weight: torch.Tensor
+    variance_epsilon: float
+    def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
+        output = ops.dropout_add_ln_fwd(
+            hidden_states.view(-1, hidden_states.shape[-1]),
+            gamma = self.weight,
+            beta = None,
+            rowscale = None,
+            colscale = None,
+            x0_subset = None,
+            z_subset = None,
+            dropout_p = 0,
+            epsilon = self.variance_epsilon,
+            rowscale_const = 1.0,
+            z_numrows = hidden_states.shape[1],
+            gen = None,
+            residual_in_fp32 = False,
+            is_rms_norm = True,
+        )
+        return output[0].view(hidden_states.shape)

build/torch212-cxx11-cu126-x86_64-linux/metadata.json ADDED Viewed

	@@ -0,0 +1,15 @@

+{
+  "name": "layer-norm",
+  "id": "_layer_norm_cuda_73ccd0c",
+  "version": 1,
+  "license": "BSD-3-Clause",
+  "python-depends": [],
+  "backend": {
+    "type": "cuda",
+    "archs": [
+      "8.0",
+      "8.9",
+      "9.0"
+    ]
+  }
+}

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

	@@ -0,0 +1,26 @@

+import torch
+import torch.nn as nn
+from ._ops import ops
+from . import layers
+def dropout_add_ln_fwd(input, gamma, beta, rowscale, colscale, x0_subset, z_subset, dropout_p, epsilon, rowscale_const, z_numrows, gen, residual_in_fp32, is_rms_norm):
+    return ops.dropout_add_ln_fwd(input, gamma, beta, rowscale, colscale, x0_subset, z_subset, dropout_p, epsilon, rowscale_const, z_numrows, gen, residual_in_fp32, is_rms_norm)
+def dropout_add_ln_bwd(dz, dx, x, mu, rsigma, gamma, rowscale, colscale, x0_subset, z_subset, dropout_p, rowscale_const, x0_numrows, has_residual, is_rms_norm):
+    return ops.dropout_add_ln_bwd(dz, dx, x, mu, rsigma, gamma, rowscale, colscale, x0_subset, z_subset, dropout_p, rowscale_const, x0_numrows, has_residual, is_rms_norm)
+def dropout_add_ln_parallel_residual_fwd(input, gamma0, beta0, gamma1, beta1, dropout_p, epsilon, gen, residual_in_fp32, is_rms_norm):
+    return ops.dropout_add_ln_parallel_residual_fwd(input, gamma0, beta0, gamma1, beta1, dropout_p, epsilon, gen, residual_in_fp32, is_rms_norm)
+def dropout_add_ln_parallel_residual_bwd(dz0, dz1, dx, x, mu, rsigma, gamma0, gamma1, dropout_p, has_x1, has_residual, is_rms_norm):
+    return ops.dropout_add_ln_parallel_residual_bwd(dz0, dz1, dx, x, mu, rsigma, gamma0, gamma1, dropout_p, has_x1, has_residual, is_rms_norm)
+__all__ = [
+    "layers",
+    "dropout_add_ln_fwd",
+    "dropout_add_ln_bwd",
+    "dropout_add_ln_parallel_residual_fwd",
+    "dropout_add_ln_parallel_residual_bwd",
+]

build/torch212-cxx11-cu130-x86_64-linux/_layer_norm_cuda_73ccd0c.abi3.so ADDED Viewed

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

+version https://git-lfs.github.com/spec/v1
+oid sha256:da280ce065bfa6679da65d46168975cbb0a52d90c9d175ac77deb228ca20756a
+size 1238392592

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

	@@ -0,0 +1,9 @@

+import torch
+from . import _layer_norm_cuda_73ccd0c
+ops = torch.ops._layer_norm_cuda_73ccd0c
+def add_op_namespace_prefix(op_name: str):
+    """
+    Prefix op by namespace.
+    """
+    return f"_layer_norm_cuda_73ccd0c::{op_name}"

build/torch212-cxx11-cu130-x86_64-linux/layer_norm/__init__.py ADDED Viewed

	@@ -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/layers.py ADDED Viewed

	@@ -0,0 +1,51 @@

+import torch
+import torch.nn as nn
+from ._ops import ops
+class LayerNorm(nn.Module):
+    weight: torch.Tensor
+    variance_epsilon: float
+    def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
+        output = ops.dropout_add_ln_fwd(
+            hidden_states.view(-1, hidden_states.shape[-1]),
+            gamma = self.weight,
+            beta = None,
+            rowscale = None,
+            colscale = None,
+            x0_subset = None,
+            z_subset = None,
+            dropout_p = 0,
+            epsilon = self.variance_epsilon,
+            rowscale_const = 1.0,
+            z_numrows = hidden_states.shape[1],
+            gen = None,
+            residual_in_fp32 = False,
+            is_rms_norm = False,
+        )
+        return output[0].view(hidden_states.shape)
+class LlamaRMSNorm(nn.Module):
+    weight: torch.Tensor
+    variance_epsilon: float
+    def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
+        output = ops.dropout_add_ln_fwd(
+            hidden_states.view(-1, hidden_states.shape[-1]),
+            gamma = self.weight,
+            beta = None,
+            rowscale = None,
+            colscale = None,
+            x0_subset = None,
+            z_subset = None,
+            dropout_p = 0,
+            epsilon = self.variance_epsilon,
+            rowscale_const = 1.0,
+            z_numrows = hidden_states.shape[1],
+            gen = None,
+            residual_in_fp32 = False,
+            is_rms_norm = True,
+        )
+        return output[0].view(hidden_states.shape)

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

	@@ -0,0 +1,17 @@

+{
+  "name": "layer-norm",
+  "id": "_layer_norm_cuda_73ccd0c",
+  "version": 1,
+  "license": "BSD-3-Clause",
+  "python-depends": [],
+  "backend": {
+    "type": "cuda",
+    "archs": [
+      "10.0",
+      "12.0",
+      "8.0",
+      "8.9",
+      "9.0"
+    ]
+  }
+}

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

	@@ -0,0 +1,26 @@

+import torch
+import torch.nn as nn
+from ._ops import ops
+from . import layers
+def dropout_add_ln_fwd(input, gamma, beta, rowscale, colscale, x0_subset, z_subset, dropout_p, epsilon, rowscale_const, z_numrows, gen, residual_in_fp32, is_rms_norm):
+    return ops.dropout_add_ln_fwd(input, gamma, beta, rowscale, colscale, x0_subset, z_subset, dropout_p, epsilon, rowscale_const, z_numrows, gen, residual_in_fp32, is_rms_norm)
+def dropout_add_ln_bwd(dz, dx, x, mu, rsigma, gamma, rowscale, colscale, x0_subset, z_subset, dropout_p, rowscale_const, x0_numrows, has_residual, is_rms_norm):
+    return ops.dropout_add_ln_bwd(dz, dx, x, mu, rsigma, gamma, rowscale, colscale, x0_subset, z_subset, dropout_p, rowscale_const, x0_numrows, has_residual, is_rms_norm)
+def dropout_add_ln_parallel_residual_fwd(input, gamma0, beta0, gamma1, beta1, dropout_p, epsilon, gen, residual_in_fp32, is_rms_norm):
+    return ops.dropout_add_ln_parallel_residual_fwd(input, gamma0, beta0, gamma1, beta1, dropout_p, epsilon, gen, residual_in_fp32, is_rms_norm)
+def dropout_add_ln_parallel_residual_bwd(dz0, dz1, dx, x, mu, rsigma, gamma0, gamma1, dropout_p, has_x1, has_residual, is_rms_norm):
+    return ops.dropout_add_ln_parallel_residual_bwd(dz0, dz1, dx, x, mu, rsigma, gamma0, gamma1, dropout_p, has_x1, has_residual, is_rms_norm)
+__all__ = [
+    "layers",
+    "dropout_add_ln_fwd",
+    "dropout_add_ln_bwd",
+    "dropout_add_ln_parallel_residual_fwd",
+    "dropout_add_ln_parallel_residual_bwd",
+]

build/torch212-cxx11-cu132-x86_64-linux/_layer_norm_cuda_73ccd0c.abi3.so ADDED Viewed

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

+version https://git-lfs.github.com/spec/v1
+oid sha256:16f936b4a001754ef5d21ca0a7d2bfb2fa8a25f713fdd166f8b1a47e524c588d
+size 1222008944

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

	@@ -0,0 +1,9 @@

+import torch
+from . import _layer_norm_cuda_73ccd0c
+ops = torch.ops._layer_norm_cuda_73ccd0c
+def add_op_namespace_prefix(op_name: str):
+    """
+    Prefix op by namespace.
+    """
+    return f"_layer_norm_cuda_73ccd0c::{op_name}"

build/torch212-cxx11-cu132-x86_64-linux/layer_norm/__init__.py ADDED Viewed

	@@ -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/layers.py ADDED Viewed

	@@ -0,0 +1,51 @@

+import torch
+import torch.nn as nn
+from ._ops import ops
+class LayerNorm(nn.Module):
+    weight: torch.Tensor
+    variance_epsilon: float
+    def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
+        output = ops.dropout_add_ln_fwd(
+            hidden_states.view(-1, hidden_states.shape[-1]),
+            gamma = self.weight,
+            beta = None,
+            rowscale = None,
+            colscale = None,
+            x0_subset = None,
+            z_subset = None,
+            dropout_p = 0,
+            epsilon = self.variance_epsilon,
+            rowscale_const = 1.0,
+            z_numrows = hidden_states.shape[1],
+            gen = None,
+            residual_in_fp32 = False,
+            is_rms_norm = False,
+        )
+        return output[0].view(hidden_states.shape)
+class LlamaRMSNorm(nn.Module):
+    weight: torch.Tensor
+    variance_epsilon: float
+    def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
+        output = ops.dropout_add_ln_fwd(
+            hidden_states.view(-1, hidden_states.shape[-1]),
+            gamma = self.weight,
+            beta = None,
+            rowscale = None,
+            colscale = None,
+            x0_subset = None,
+            z_subset = None,
+            dropout_p = 0,
+            epsilon = self.variance_epsilon,
+            rowscale_const = 1.0,
+            z_numrows = hidden_states.shape[1],
+            gen = None,
+            residual_in_fp32 = False,
+            is_rms_norm = True,
+        )
+        return output[0].view(hidden_states.shape)

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

	@@ -0,0 +1,17 @@

+{
+  "name": "layer-norm",
+  "id": "_layer_norm_cuda_73ccd0c",
+  "version": 1,
+  "license": "BSD-3-Clause",
+  "python-depends": [],
+  "backend": {
+    "type": "cuda",
+    "archs": [
+      "10.0",
+      "12.0",
+      "8.0",
+      "8.9",
+      "9.0"
+    ]
+  }
+}