Kernels:

kernels-community
/

activation

Trusted publisher

Kernel card Files Files and versions

xet

Community

kernels-bot commited on Jul 9

Commit

6824b2c

verified ·

1 Parent(s): daadf59

Uploaded using `kernel-builder`.

Browse files

Files changed (28) hide show

build/torch-stable-abi210-cu126-aarch64-linux/__init__.py +75 -0
build/torch-stable-abi210-cu126-aarch64-linux/_activation_cuda_2f7c452.abi3.so +3 -0
build/torch-stable-abi210-cu126-aarch64-linux/_ops.py +9 -0
build/torch-stable-abi210-cu126-aarch64-linux/activation/__init__.py +26 -0
build/torch-stable-abi210-cu126-aarch64-linux/layers.py +201 -0
build/torch-stable-abi210-cu126-aarch64-linux/metadata.json +41 -0
build/torch-stable-abi210-cu126-aarch64-linux/metadata.json.sigstore +1 -0
build/torch-stable-abi210-cu128-aarch64-linux/__init__.py +75 -0
build/torch-stable-abi210-cu128-aarch64-linux/_activation_cuda_2f7c452.abi3.so +3 -0
build/torch-stable-abi210-cu128-aarch64-linux/_ops.py +9 -0
build/torch-stable-abi210-cu128-aarch64-linux/activation/__init__.py +26 -0
build/torch-stable-abi210-cu128-aarch64-linux/layers.py +201 -0
build/torch-stable-abi210-cu128-aarch64-linux/metadata.json +44 -0
build/torch-stable-abi210-cu128-aarch64-linux/metadata.json.sigstore +1 -0
build/torch-stable-abi210-cu130-aarch64-linux/__init__.py +75 -0
build/torch-stable-abi210-cu130-aarch64-linux/_activation_cuda_2f7c452.abi3.so +3 -0
build/torch-stable-abi210-cu130-aarch64-linux/_ops.py +9 -0
build/torch-stable-abi210-cu130-aarch64-linux/activation/__init__.py +26 -0
build/torch-stable-abi210-cu130-aarch64-linux/layers.py +201 -0
build/torch-stable-abi210-cu130-aarch64-linux/metadata.json +43 -0
build/torch-stable-abi210-cu130-aarch64-linux/metadata.json.sigstore +1 -0
build/torch-stable-abi210-cu132-aarch64-linux/__init__.py +75 -0
build/torch-stable-abi210-cu132-aarch64-linux/_activation_cuda_2f7c452.abi3.so +3 -0
build/torch-stable-abi210-cu132-aarch64-linux/_ops.py +9 -0
build/torch-stable-abi210-cu132-aarch64-linux/activation/__init__.py +26 -0
build/torch-stable-abi210-cu132-aarch64-linux/layers.py +201 -0
build/torch-stable-abi210-cu132-aarch64-linux/metadata.json +43 -0
build/torch-stable-abi210-cu132-aarch64-linux/metadata.json.sigstore +1 -0

build/torch-stable-abi210-cu126-aarch64-linux/__init__.py ADDED Viewed

	@@ -0,0 +1,75 @@

+import torch
+from ._ops import ops
+from . import layers
+def silu_and_mul(out: torch.Tensor, x: torch.Tensor) -> torch.Tensor:
+    ops.silu_and_mul(out, x)
+    return out
+def mul_and_silu(out: torch.Tensor, x: torch.Tensor) -> torch.Tensor:
+    ops.mul_and_silu(out, x)
+    return out
+def gelu_and_mul(out: torch.Tensor, x: torch.Tensor) -> torch.Tensor:
+    ops.gelu_and_mul(out, x)
+    return out
+def gelu_tanh_and_mul(out: torch.Tensor, x: torch.Tensor) -> torch.Tensor:
+    ops.gelu_tanh_and_mul(out, x)
+    return out
+def fatrelu_and_mul(out: torch.Tensor, x: torch.Tensor, threshold: float = 0.0) -> torch.Tensor:
+    ops.fatrelu_and_mul(out, x, threshold)
+    return out
+def gelu(out: torch.Tensor, x: torch.Tensor) -> torch.Tensor:
+    ops.gelu(out, x)
+    return out
+def silu(out: torch.Tensor, x: torch.Tensor) -> torch.Tensor:
+    ops.silu(out, x)
+    return out
+def gelu_tanh(out: torch.Tensor, x: torch.Tensor) -> torch.Tensor:
+    ops.gelu_tanh(out, x)
+    return out
+def gelu_fast(out: torch.Tensor, x: torch.Tensor) -> torch.Tensor:
+    ops.gelu_fast(out, x)
+    return out
+def gelu_new(out: torch.Tensor, x: torch.Tensor) -> torch.Tensor:
+    ops.gelu_new(out, x)
+    return out
+def gelu_quick(out: torch.Tensor, x: torch.Tensor) -> torch.Tensor:
+    ops.gelu_quick(out, x)
+    return out
+__all__ = [
+    "silu_and_mul",
+    "mul_and_silu",
+    "gelu_and_mul",
+    "gelu_tanh_and_mul",
+    "fatrelu_and_mul",
+    "gelu_fast",
+    "gelu_new",
+    "gelu_quick",
+    "gelu_tanh",
+    "silu",
+    "gelu",
+    "layers",
+]

build/torch-stable-abi210-cu126-aarch64-linux/_activation_cuda_2f7c452.abi3.so ADDED Viewed

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

+version https://git-lfs.github.com/spec/v1
+oid sha256:ef2c3eb198b5a665ea644ba7aac650f1955de3047adb52e63e0bc9c435c1dcf7
+size 1269632

build/torch-stable-abi210-cu126-aarch64-linux/_ops.py ADDED Viewed

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

+import torch
+from . import _activation_cuda_2f7c452
+ops = torch.ops._activation_cuda_2f7c452
+def add_op_namespace_prefix(op_name: str):
+    """
+    Prefix op by namespace.
+    """
+    return f"_activation_cuda_2f7c452::{op_name}"

build/torch-stable-abi210-cu126-aarch64-linux/activation/__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/torch-stable-abi210-cu126-aarch64-linux/layers.py ADDED Viewed

	@@ -0,0 +1,201 @@

+import torch
+import torch.nn as nn
+from ._ops import ops
+class SiluAndMul(nn.Module):
+    """An activation function for SwiGLU.
+    The function computes x -> silu(x[:d]) * x[d:] where d = x.shape[-1] // 2.
+    Shapes:
+        x: (num_tokens, 2 * d) or (batch_size, seq_len, 2 * d)
+        return: (num_tokens, d) or (batch_size, seq_len, d)
+    """
+    can_torch_compile: bool = True
+    def forward(self, x: torch.Tensor):
+        if not x.is_contiguous():
+            x = x.contiguous()
+        d = x.shape[-1] // 2
+        output_shape = x.shape[:-1] + (d,)
+        out = torch.empty(output_shape, dtype=x.dtype, device=x.device)
+        ops.silu_and_mul(out, x)
+        return out
+class Silu(nn.Module):
+    """An activation function for SiLU.
+    The function computes x -> silu(x).
+    Shapes:
+        x: (num_tokens, d) or (batch_size, seq_len, d)
+        return: (num_tokens, d) or (batch_size, seq_len, d)
+    """
+    can_torch_compile: bool = True
+    def forward(self, x: torch.Tensor):
+        if not x.is_contiguous():
+            x = x.contiguous()
+        out = torch.empty_like(x)
+        ops.silu(out, x)
+        return out
+class Gelu(nn.Module):
+    """An activation function for GELU.
+    The function computes x -> gelu(x).
+    Shapes:
+        x: (num_tokens, d) or (batch_size, seq_len, d)
+        return: (num_tokens, d) or (batch_size, seq_len, d)
+    """
+    can_torch_compile: bool = True
+    def forward(self, x: torch.Tensor):
+        if not x.is_contiguous():
+            x = x.contiguous()
+        out = torch.empty_like(x)
+        ops.gelu(out, x)
+        return out
+class GeluTanh(nn.Module):
+    """An activation function for GELU with `tanh` approximation.
+    The function computes x -> gelu_tanh(x).
+    Shapes:
+        x: (num_tokens, d) or (batch_size, seq_len, d)
+        return: (num_tokens, d) or (batch_size, seq_len, d)
+    """
+    can_torch_compile: bool = True
+    def forward(self, x: torch.Tensor):
+        if not x.is_contiguous():
+            x = x.contiguous()
+        out = torch.empty_like(x)
+        ops.gelu_tanh(out, x)
+        return out
+class MulAndSilu(nn.Module):
+    """An activation function for SwiGLU.
+    The function computes x -> x[:d] * silu(x[d:]) where d = x.shape[-1] // 2.
+    Shapes:
+        x: (num_tokens, 2 * d) or (batch_size, seq_len, 2 * d)
+        return: (num_tokens, d) or (batch_size, seq_len, d)
+    """
+    can_torch_compile: bool = True
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        if not x.is_contiguous():
+            x = x.contiguous()
+        d = x.shape[-1] // 2
+        output_shape = x.shape[:-1] + (d,)
+        out = torch.empty(output_shape, dtype=x.dtype, device=x.device)
+        ops.mul_and_silu(out, x)
+        return out
+class GeluAndMul(nn.Module):
+    """An activation function for GeGLU.
+    The function computes x -> GELU(x[:d]) * x[d:] where d = x.shape[-1] // 2.
+    Shapes:
+        x: (batch_size, seq_len, 2 * d) or (num_tokens, 2 * d)
+        return: (batch_size, seq_len, d) or (num_tokens, d)
+    """
+    can_torch_compile: bool = True
+    def forward(self, x: torch.Tensor):
+        if not x.is_contiguous():
+            x = x.contiguous()
+        d = x.shape[-1] // 2
+        output_shape = x.shape[:-1] + (d,)
+        out = torch.empty(output_shape, dtype=x.dtype, device=x.device)
+        ops.gelu_and_mul(out, x)
+        return out
+class GeluTanhAndMul(nn.Module):
+    can_torch_compile: bool = True
+    def forward(self, x: torch.Tensor):
+        if not x.is_contiguous():
+            x = x.contiguous()
+        d = x.shape[-1] // 2
+        output_shape = x.shape[:-1] + (d,)
+        out = torch.empty(output_shape, dtype=x.dtype, device=x.device)
+        ops.gelu_tanh_and_mul(out, x)
+        return out
+class FatreluAndMul(nn.Module):
+    """An activation function for FATReLU.
+    The function computes x -> FATReLU(x[:d]) * x[d:] where
+    d = x.shape[-1] // 2.
+    This is used in openbmb/MiniCPM-S-1B-sft.
+    Shapes:
+        x: (num_tokens, 2 * d) or (batch_size, seq_len, 2 * d)
+        return: (num_tokens, d) or (batch_size, seq_len, d)
+    """
+    can_torch_compile: bool = True
+    def __init__(self, threshold: float = 0.0):
+        super().__init__()
+        self.threshold = threshold
+    def forward(self, x: torch.Tensor):
+        if not x.is_contiguous():
+            x = x.contiguous()
+        d = x.shape[-1] // 2
+        output_shape = x.shape[:-1] + (d,)
+        out = torch.empty(output_shape, dtype=x.dtype, device=x.device)
+        ops.fatrelu_and_mul(out, x, self.threshold)
+        return out
+class FastGELU(nn.Module):
+    can_torch_compile: bool = True
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        if not x.is_contiguous():
+            x = x.contiguous()
+        out = torch.empty_like(x)
+        ops.gelu_fast(out, x)
+        return out
+class NewGELU(nn.Module):
+    can_torch_compile: bool = True
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        if not x.is_contiguous():
+            x = x.contiguous()
+        out = torch.empty_like(x)
+        ops.gelu_new(out, x)
+        return out
+class QuickGELU(nn.Module):
+    can_torch_compile: bool = True
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        if not x.is_contiguous():
+            x = x.contiguous()
+        out = torch.empty_like(x)
+        ops.gelu_quick(out, x)
+        return out

build/torch-stable-abi210-cu126-aarch64-linux/metadata.json ADDED Viewed

	@@ -0,0 +1,41 @@

+{
+  "name": "activation",
+  "id": "_activation_cuda_2f7c452",
+  "version": 1,
+  "license": "Apache-2.0",
+  "python-depends": [],
+  "backend": {
+    "type": "cuda",
+    "archs": [
+      "7.0",
+      "7.2",
+      "7.5",
+      "8.0",
+      "8.6",
+      "8.7",
+      "8.9",
+      "9.0+PTX"
+    ]
+  },
+  "digest": {
+    "algorithm": "sha256",
+    "files": {
+      "__init__.py": "NQdhY7p9BPoMw0NQodMKFTPSQXBlrsX3yUdP8MEHyH0=",
+      "_activation_cuda_2f7c452.abi3.so": "7yw+sZi1pmXqZEunqsZQ8ZVd4wR621LmPgvJxDXB3Pc=",
+      "_ops.py": "q9xZtwPjdsRciOCA5aQFDtsTQnV9SNcvAC74ivbShuM=",
+      "activation/__init__.py": "DFYPlrhXwYjEqCl/8n0SmWGZV8NFml5DPhMjKfv98GY=",
+      "layers.py": "wWX07+W0BNFA7RyCE0jYH7aa2U3B8aiiIpu/Q4xsE0w="
+    }
+  },
+  "provenance": {
+    "kernel-builder": {
+      "version": "0.17.0-dev0",
+      "sha": "81580bb92577f2f7228661ca2a221fc052375709",
+      "dirty": false
+    },
+    "kernel": {
+      "sha": "2f7c452801641012cf6aa9bdf042431457024033",
+      "dirty": false
+    }
+  }
+}

build/torch-stable-abi210-cu126-aarch64-linux/metadata.json.sigstore ADDED Viewed

	@@ -0,0 +1 @@

+ {"mediaType":"application/vnd.dev.sigstore.bundle.v0.3+json", "verificationMaterial":{"certificate":{"rawBytes":"MIIHSzCCBtGgAwIBAgIUWSLB8pFWUwWaHXhGDZ3R8RhNi2wwCgYIKoZIzj0EAwMwNzEVMBMGA1UEChMMc2lnc3RvcmUuZGV2MR4wHAYDVQQDExVzaWdzdG9yZS1pbnRlcm1lZGlhdGUwHhcNMjYwNzA5MTU0OTQ0WhcNMjYwNzA5MTU1OTQ0WjAAMFkwEwYHKoZIzj0CAQYIKoZIzj0DAQcDQgAE3Q0LUHl8Q/dpD61n2CUkQYAvMuAC81UdHTNTy8oT9TsA2PCuLhBZ6lQhcN20FUps1WTv4UEQCvrfVdzCAsQOKqOCBfAwggXsMA4GA1UdDwEB/wQEAwIHgDATBgNVHSUEDDAKBggrBgEFBQcDAzAdBgNVHQ4EFgQUZ0zSs0EZ59HIag73bqHIUj1BmsgwHwYDVR0jBBgwFoAU39Ppz1YkEZb5qNjpKFWixi4YZD8wawYDVR0RAQH/BGEwX4ZdaHR0cHM6Ly9naXRodWIuY29tL2h1Z2dpbmdmYWNlL2tlcm5lbHMtY29tbXVuaXR5Ly5naXRodWIvd29ya2Zsb3dzL2J1aWxkLnlhbWxAcmVmcy9oZWFkcy9tYWluMDkGCisGAQQBg78wAQEEK2h0dHBzOi8vdG9rZW4uYWN0aW9ucy5naXRodWJ1c2VyY29udGVudC5jb20wHwYKKwYBBAGDvzABAgQRd29ya2Zsb3dfZGlzcGF0Y2gwNgYKKwYBBAGDvzABAwQoMmY3YzQ1MjgwMTY0MTAxMmNmNmFhOWJkZjA0MjQzMTQ1NzAyNDAzMzATBgorBgEEAYO/MAEEBAVCdWlsZDArBgorBgEEAYO/MAEFBB1odWdnaW5nZmFjZS9rZXJuZWxzLWNvbW11bml0eTAdBgorBgEEAYO/MAEGBA9yZWZzL2hlYWRzL21haW4wOwYKKwYBBAGDvzABCAQtDCtodHRwczovL3Rva2VuLmFjdGlvbnMuZ2l0aHVidXNlcmNvbnRlbnQuY29tMG0GCisGAQQBg78wAQkEXwxdaHR0cHM6Ly9naXRodWIuY29tL2h1Z2dpbmdmYWNlL2tlcm5lbHMtY29tbXVuaXR5Ly5naXRodWIvd29ya2Zsb3dzL2J1aWxkLnlhbWxAcmVmcy9oZWFkcy9tYWluMDgGCisGAQQBg78wAQoEKgwoMmY3YzQ1MjgwMTY0MTAxMmNmNmFhOWJkZjA0MjQzMTQ1NzAyNDAzMzAbBgorBgEEAYO/MAELBA0MC3NlbGYtaG9zdGVkMEAGCisGAQQBg78wAQwEMgwwaHR0cHM6Ly9naXRodWIuY29tL2h1Z2dpbmdmYWNlL2tlcm5lbHMtY29tbXVuaXR5MDgGCisGAQQBg78wAQ0EKgwoMmY3YzQ1MjgwMTY0MTAxMmNmNmFhOWJkZjA0MjQzMTQ1NzAyNDAzMzAfBgorBgEEAYO/MAEOBBEMD3JlZnMvaGVhZHMvbWFpbjAaBgorBgEEAYO/MAEPBAwMCjEwNzE0NzU1MjkwLgYKKwYBBAGDvzABEAQgDB5odHRwczovL2dpdGh1Yi5jb20vaHVnZ2luZ2ZhY2UwGAYKKwYBBAGDvzABEQQKDAgyNTcyMDc0MzBtBgorBgEEAYO/MAESBF8MXWh0dHBzOi8vZ2l0aHViLmNvbS9odWdnaW5nZmFjZS9rZXJuZWxzLWNvbW11bml0eS8uZ2l0aHViL3dvcmtmbG93cy9idWlsZC55YW1sQHJlZnMvaGVhZHMvbWFpbjA4BgorBgEEAYO/MAETBCoMKDJmN2M0NTI4MDE2NDEwMTJjZjZhYTliZGYwNDI0MzE0NTcwMjQwMzMwIQYKKwYBBAGDvzABFAQTDBF3b3JrZmxvd19kaXNwYXRjaDBkBgorBgEEAYO/MAEVBFYMVGh0dHBzOi8vZ2l0aHViLmNvbS9odWdnaW5nZmFjZS9rZXJuZWxzLWNvbW11bml0eS9hY3Rpb25zL3J1bnMvMjkwMzA3NzExMDMvYXR0ZW1wdHMvMTAWBgorBgEEAYO/MAEWBAgMBnB1YmxpYzBGBgorBgEEAYO/MAEYBDgMNnJlcG86aHVnZ2luZ2ZhY2Uva2VybmVscy1jb21tdW5pdHk6cmVmOnJlZnMvaGVhZHMvbWFpbjCBigYKKwYBBAHWeQIEAgR8BHoAeAB2AN09MGrGxxEyYxkeHJlnNwKiSl643jyt/4eKcoAvKe6OAAABn0eSFw8AAAQDAEcwRQIgCPNxq5BTCEvYNB5Yd1eFJxZe+S1wnTU0duaEd8czyyACIQC7xLCjz3usA58QG4lqIAKMqEgfxzFTBu2l/FT6+/RaLTAKBggqhkjOPQQDAwNoADBlAjBijZ0ZYeOpnHE8FeF1fjw6FxjXf996xYfgJ9xD3zLISgDwgPnmXeQJeG2cvNUQKP4CMQCaFuDZHFUv9N8HAkmcoDUntzDUFmAAU6rEA9XJo+pJ9SYJNksOOftMkJkCTbbrzTk="}, "tlogEntries":[{"logIndex":"2129276360", "logId":{"keyId":"wNI9atQGlz+VWfO6LRygH4QUfY/8W4RFwiT5i5WRgB0="}, "kindVersion":{"kind":"hashedrekord", "version":"0.0.1"}, "integratedTime":"1783612184", "inclusionPromise":{"signedEntryTimestamp":"MEYCIQCAG+2qd1y6L20nKwKjZ5gK9gBfhidzwOgmFlLJOUxPTAIhAOSWhZZbpy5EKp8tEnE0FNRKRUjZBjwyaRMLsEq8OZ2+"}, "inclusionProof":{"logIndex":"2007372098", "rootHash":"rAKgnWZPr5tDfgQYbD/tDv6hEP2z32lecojkS/lf+Kw=", "treeSize":"2007372203", "hashes":["11kc0w4vx871JuhWd77UJtrhs9IrIh061lA1A4Q3L1M=", "dG9TIud97f1PwHDA8+PZ6O2l5SbAwyrL3vEPmvxg+BI=", "y7/r5j6sd+lMIyxP1yCNjlRRvyR69hE8F3QcjicwAf8=", "O2y/ucAN1JhlXc6gUHDnY1vWSd+yKQH5r7nj0UNIq+c=", "jLUKBDwQTkxc3n0lYkMG6cokfw241MTMSEjhEjXfJeE=", "2f+TLUtk3VTZLdu97iTZEzv6NVa4fptRmlr1vt3XkN0=", "cVSauhvBQaIgdstH+cI20VNVCxXCGkMGtf5sbMMY6DY=", "EiAa30gkrCiVgCEkDPQGRjaC553qLkiEl8xMQ9bUiew=", "bnulwvz3BzvG+I9Vi/SUXkNNoZVe0J6+bGWiykXUvl4=", "6q+LpbK/Ih183cA9KuL8+fgXqZf4ltUk/2rp524ZFuI=", "CCu7OkQW6+npAh6Tygn9JYdKku6AUboDosZP3sI7yKQ=", "uwnh4tLrO/3+ywQ63CnFW6mU2V+y+Db8CobG3g6S0Hk=", "Kezfh1XcoOp9KAwr7OWyF8C7cr/mZBPGqKHTdO0BnDs=", "LD5Ws0n7JWm8O7qGvHNSZug32ByFM1WblCpsDJh5M6o=", "f8knrc4LOu/X8scT5SBvEgFJSd7vZZ0LzSaC0Z3lm/U=", "WajL3ZhY6ZX9meV3VTH9TkrbM31OF4o6+7foL0Uf6PI=", "cyKMKBvHkQtwb2xR5svh+YppANOVdn7X8t4DuKMxICA=", "+/VZ56MsIPxMiyLAodzKXo5TEWdQp36z89qLhpzloAo=", "daxmZaajRpZV+JxHiOYZhJBiSKN5ucqjh2WnGbHhirw=", "DOCeoSMovIvLExkhIvisow9AuNXgeWs4ECkyR6EcqYU="], "checkpoint":{"envelope":"rekor.sigstore.dev - 1193050959916656506\n2007372203\nrAKgnWZPr5tDfgQYbD/tDv6hEP2z32lecojkS/lf+Kw=\n\n— rekor.sigstore.dev wNI9ajBEAiBLnnP9qGB3ZIj2vvAuH8e5nYgoxwJFUCPLn/FyIEhiMQIgHwIVNTAbnEQ417uGoOfvN43IT3JM3U4WL2k1BSOwj7A=\n"}}, "canonicalizedBody":"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"}], "timestampVerificationData":{"rfc3161Timestamps":[{"signedTimestamp":"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"}]}}, "messageSignature":{"messageDigest":{"algorithm":"SHA2_256", "digest":"IlhkfrRwf5/puBJNmMGcOaaMDB/Ro9LklyKQdvv6YHg="}, "signature":"MEUCIQDMiXpSgRJWfY+Luj/x6U14A+9HSfSjseBp3/xieZtXUwIgLKlqZSLquGngHBE2hCdLsKWtFVcmBzuGxs8kY4s1kbI="}}

build/torch-stable-abi210-cu128-aarch64-linux/__init__.py ADDED Viewed

	@@ -0,0 +1,75 @@

+import torch
+from ._ops import ops
+from . import layers
+def silu_and_mul(out: torch.Tensor, x: torch.Tensor) -> torch.Tensor:
+    ops.silu_and_mul(out, x)
+    return out
+def mul_and_silu(out: torch.Tensor, x: torch.Tensor) -> torch.Tensor:
+    ops.mul_and_silu(out, x)
+    return out
+def gelu_and_mul(out: torch.Tensor, x: torch.Tensor) -> torch.Tensor:
+    ops.gelu_and_mul(out, x)
+    return out
+def gelu_tanh_and_mul(out: torch.Tensor, x: torch.Tensor) -> torch.Tensor:
+    ops.gelu_tanh_and_mul(out, x)
+    return out
+def fatrelu_and_mul(out: torch.Tensor, x: torch.Tensor, threshold: float = 0.0) -> torch.Tensor:
+    ops.fatrelu_and_mul(out, x, threshold)
+    return out
+def gelu(out: torch.Tensor, x: torch.Tensor) -> torch.Tensor:
+    ops.gelu(out, x)
+    return out
+def silu(out: torch.Tensor, x: torch.Tensor) -> torch.Tensor:
+    ops.silu(out, x)
+    return out
+def gelu_tanh(out: torch.Tensor, x: torch.Tensor) -> torch.Tensor:
+    ops.gelu_tanh(out, x)
+    return out
+def gelu_fast(out: torch.Tensor, x: torch.Tensor) -> torch.Tensor:
+    ops.gelu_fast(out, x)
+    return out
+def gelu_new(out: torch.Tensor, x: torch.Tensor) -> torch.Tensor:
+    ops.gelu_new(out, x)
+    return out
+def gelu_quick(out: torch.Tensor, x: torch.Tensor) -> torch.Tensor:
+    ops.gelu_quick(out, x)
+    return out
+__all__ = [
+    "silu_and_mul",
+    "mul_and_silu",
+    "gelu_and_mul",
+    "gelu_tanh_and_mul",
+    "fatrelu_and_mul",
+    "gelu_fast",
+    "gelu_new",
+    "gelu_quick",
+    "gelu_tanh",
+    "silu",
+    "gelu",
+    "layers",
+]

build/torch-stable-abi210-cu128-aarch64-linux/_activation_cuda_2f7c452.abi3.so ADDED Viewed

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

+version https://git-lfs.github.com/spec/v1
+oid sha256:ed51ba1cc66a43004c27a52997846193dcc77c44b8ecf0909b47084d69528c8d
+size 2449328

build/torch-stable-abi210-cu128-aarch64-linux/_ops.py ADDED Viewed

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

+import torch
+from . import _activation_cuda_2f7c452
+ops = torch.ops._activation_cuda_2f7c452
+def add_op_namespace_prefix(op_name: str):
+    """
+    Prefix op by namespace.
+    """
+    return f"_activation_cuda_2f7c452::{op_name}"

build/torch-stable-abi210-cu128-aarch64-linux/activation/__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/torch-stable-abi210-cu128-aarch64-linux/layers.py ADDED Viewed

	@@ -0,0 +1,201 @@

+import torch
+import torch.nn as nn
+from ._ops import ops
+class SiluAndMul(nn.Module):
+    """An activation function for SwiGLU.
+    The function computes x -> silu(x[:d]) * x[d:] where d = x.shape[-1] // 2.
+    Shapes:
+        x: (num_tokens, 2 * d) or (batch_size, seq_len, 2 * d)
+        return: (num_tokens, d) or (batch_size, seq_len, d)
+    """
+    can_torch_compile: bool = True
+    def forward(self, x: torch.Tensor):
+        if not x.is_contiguous():
+            x = x.contiguous()
+        d = x.shape[-1] // 2
+        output_shape = x.shape[:-1] + (d,)
+        out = torch.empty(output_shape, dtype=x.dtype, device=x.device)
+        ops.silu_and_mul(out, x)
+        return out
+class Silu(nn.Module):
+    """An activation function for SiLU.
+    The function computes x -> silu(x).
+    Shapes:
+        x: (num_tokens, d) or (batch_size, seq_len, d)
+        return: (num_tokens, d) or (batch_size, seq_len, d)
+    """
+    can_torch_compile: bool = True
+    def forward(self, x: torch.Tensor):
+        if not x.is_contiguous():
+            x = x.contiguous()
+        out = torch.empty_like(x)
+        ops.silu(out, x)
+        return out
+class Gelu(nn.Module):
+    """An activation function for GELU.
+    The function computes x -> gelu(x).
+    Shapes:
+        x: (num_tokens, d) or (batch_size, seq_len, d)
+        return: (num_tokens, d) or (batch_size, seq_len, d)
+    """
+    can_torch_compile: bool = True
+    def forward(self, x: torch.Tensor):
+        if not x.is_contiguous():
+            x = x.contiguous()
+        out = torch.empty_like(x)
+        ops.gelu(out, x)
+        return out
+class GeluTanh(nn.Module):
+    """An activation function for GELU with `tanh` approximation.
+    The function computes x -> gelu_tanh(x).
+    Shapes:
+        x: (num_tokens, d) or (batch_size, seq_len, d)
+        return: (num_tokens, d) or (batch_size, seq_len, d)
+    """
+    can_torch_compile: bool = True
+    def forward(self, x: torch.Tensor):
+        if not x.is_contiguous():
+            x = x.contiguous()
+        out = torch.empty_like(x)
+        ops.gelu_tanh(out, x)
+        return out
+class MulAndSilu(nn.Module):
+    """An activation function for SwiGLU.
+    The function computes x -> x[:d] * silu(x[d:]) where d = x.shape[-1] // 2.
+    Shapes:
+        x: (num_tokens, 2 * d) or (batch_size, seq_len, 2 * d)
+        return: (num_tokens, d) or (batch_size, seq_len, d)
+    """
+    can_torch_compile: bool = True
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        if not x.is_contiguous():
+            x = x.contiguous()
+        d = x.shape[-1] // 2
+        output_shape = x.shape[:-1] + (d,)
+        out = torch.empty(output_shape, dtype=x.dtype, device=x.device)
+        ops.mul_and_silu(out, x)
+        return out
+class GeluAndMul(nn.Module):
+    """An activation function for GeGLU.
+    The function computes x -> GELU(x[:d]) * x[d:] where d = x.shape[-1] // 2.
+    Shapes:
+        x: (batch_size, seq_len, 2 * d) or (num_tokens, 2 * d)
+        return: (batch_size, seq_len, d) or (num_tokens, d)
+    """
+    can_torch_compile: bool = True
+    def forward(self, x: torch.Tensor):
+        if not x.is_contiguous():
+            x = x.contiguous()
+        d = x.shape[-1] // 2
+        output_shape = x.shape[:-1] + (d,)
+        out = torch.empty(output_shape, dtype=x.dtype, device=x.device)
+        ops.gelu_and_mul(out, x)
+        return out
+class GeluTanhAndMul(nn.Module):
+    can_torch_compile: bool = True
+    def forward(self, x: torch.Tensor):
+        if not x.is_contiguous():
+            x = x.contiguous()
+        d = x.shape[-1] // 2
+        output_shape = x.shape[:-1] + (d,)
+        out = torch.empty(output_shape, dtype=x.dtype, device=x.device)
+        ops.gelu_tanh_and_mul(out, x)
+        return out
+class FatreluAndMul(nn.Module):
+    """An activation function for FATReLU.
+    The function computes x -> FATReLU(x[:d]) * x[d:] where
+    d = x.shape[-1] // 2.
+    This is used in openbmb/MiniCPM-S-1B-sft.
+    Shapes:
+        x: (num_tokens, 2 * d) or (batch_size, seq_len, 2 * d)
+        return: (num_tokens, d) or (batch_size, seq_len, d)
+    """
+    can_torch_compile: bool = True
+    def __init__(self, threshold: float = 0.0):
+        super().__init__()
+        self.threshold = threshold
+    def forward(self, x: torch.Tensor):
+        if not x.is_contiguous():
+            x = x.contiguous()
+        d = x.shape[-1] // 2
+        output_shape = x.shape[:-1] + (d,)
+        out = torch.empty(output_shape, dtype=x.dtype, device=x.device)
+        ops.fatrelu_and_mul(out, x, self.threshold)
+        return out
+class FastGELU(nn.Module):
+    can_torch_compile: bool = True
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        if not x.is_contiguous():
+            x = x.contiguous()
+        out = torch.empty_like(x)
+        ops.gelu_fast(out, x)
+        return out
+class NewGELU(nn.Module):
+    can_torch_compile: bool = True
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        if not x.is_contiguous():
+            x = x.contiguous()
+        out = torch.empty_like(x)
+        ops.gelu_new(out, x)
+        return out
+class QuickGELU(nn.Module):
+    can_torch_compile: bool = True
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        if not x.is_contiguous():
+            x = x.contiguous()
+        out = torch.empty_like(x)
+        ops.gelu_quick(out, x)
+        return out

build/torch-stable-abi210-cu128-aarch64-linux/metadata.json ADDED Viewed

	@@ -0,0 +1,44 @@

+{
+  "name": "activation",
+  "id": "_activation_cuda_2f7c452",
+  "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"
+    ]
+  },
+  "digest": {
+    "algorithm": "sha256",
+    "files": {
+      "__init__.py": "NQdhY7p9BPoMw0NQodMKFTPSQXBlrsX3yUdP8MEHyH0=",
+      "_activation_cuda_2f7c452.abi3.so": "7VG6HMZqQwBMJ6Upl4Rhk9zHfES47PCQm0cITWlSjI0=",
+      "_ops.py": "q9xZtwPjdsRciOCA5aQFDtsTQnV9SNcvAC74ivbShuM=",
+      "activation/__init__.py": "DFYPlrhXwYjEqCl/8n0SmWGZV8NFml5DPhMjKfv98GY=",
+      "layers.py": "wWX07+W0BNFA7RyCE0jYH7aa2U3B8aiiIpu/Q4xsE0w="
+    }
+  },
+  "provenance": {
+    "kernel-builder": {
+      "version": "0.17.0-dev0",
+      "sha": "81580bb92577f2f7228661ca2a221fc052375709",
+      "dirty": false
+    },
+    "kernel": {
+      "sha": "2f7c452801641012cf6aa9bdf042431457024033",
+      "dirty": false
+    }
+  }
+}

build/torch-stable-abi210-cu128-aarch64-linux/metadata.json.sigstore ADDED Viewed

	@@ -0,0 +1 @@

+ {"mediaType":"application/vnd.dev.sigstore.bundle.v0.3+json", "verificationMaterial":{"certificate":{"rawBytes":"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"}, "tlogEntries":[{"logIndex":"2129276718", "logId":{"keyId":"wNI9atQGlz+VWfO6LRygH4QUfY/8W4RFwiT5i5WRgB0="}, "kindVersion":{"kind":"hashedrekord", "version":"0.0.1"}, "integratedTime":"1783612186", "inclusionPromise":{"signedEntryTimestamp":"MEUCIQD+csGDoSfxn4MMJwRK53yA8KVHjcpD6xK6+NfZ6f+vuAIgPUvMTGalGBS/DhTBs6JhXRyQ3narB7mn1wsPJN9DdiE="}, "inclusionProof":{"logIndex":"2007372456", "rootHash":"07nhTX4vHqc0UWWW6M5syZitiZFDgFlnSyoZM2aN+vA=", "treeSize":"2007372487", "hashes":["7OHgb4MPAWXGhnwTwcoD72Xz7oHQBhX3PehTLyQMHWc=", "WuM43Q6zRGLn2I9Old3xCcToyXF3INHS2J7bzP1tsFk=", "RWH9BIJqI6Nx/1c29bvrygiVdUYn/yRxdnpN9rV5G9A=", "uQ/TUYesAvVeDUO1OKBKRIpBZOWyVl8G9aQfbwB1+nI=", "J+2INrZa+dDlrukuUoeplAjrF2OVZNMR89XNJZ92evE=", "mQ7YmzzG6U0kvpBS2G44JWLybVnZZaNorI/s5Wo0Tqc=", "oEOA8hf4OSwVpAPL8BcQMXOd5emCAvSUwqzuTKqGKJI=", "e67CUiGR0Vw/L2NdMUQMikicr3JK9TkwzG4prAunGDc=", "slqOv5gBmQhRJrilY7eSfegXrc1VoSHecwywwt4cYoA=", "6q+LpbK/Ih183cA9KuL8+fgXqZf4ltUk/2rp524ZFuI=", "CCu7OkQW6+npAh6Tygn9JYdKku6AUboDosZP3sI7yKQ=", "uwnh4tLrO/3+ywQ63CnFW6mU2V+y+Db8CobG3g6S0Hk=", "Kezfh1XcoOp9KAwr7OWyF8C7cr/mZBPGqKHTdO0BnDs=", "LD5Ws0n7JWm8O7qGvHNSZug32ByFM1WblCpsDJh5M6o=", "f8knrc4LOu/X8scT5SBvEgFJSd7vZZ0LzSaC0Z3lm/U=", "WajL3ZhY6ZX9meV3VTH9TkrbM31OF4o6+7foL0Uf6PI=", "cyKMKBvHkQtwb2xR5svh+YppANOVdn7X8t4DuKMxICA=", "+/VZ56MsIPxMiyLAodzKXo5TEWdQp36z89qLhpzloAo=", "daxmZaajRpZV+JxHiOYZhJBiSKN5ucqjh2WnGbHhirw=", "DOCeoSMovIvLExkhIvisow9AuNXgeWs4ECkyR6EcqYU="], "checkpoint":{"envelope":"rekor.sigstore.dev - 1193050959916656506\n2007372487\n07nhTX4vHqc0UWWW6M5syZitiZFDgFlnSyoZM2aN+vA=\n\n— rekor.sigstore.dev wNI9ajBEAiA+A4tS21EMMkEuITeXKyPBIBCIU6R3qCirCJGJhSuBSQIgT06d2RxvXPJ69rNlGga6P6q5xMKZoIR4x5zrIoMsedw=\n"}}, "canonicalizedBody":"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"}], "timestampVerificationData":{"rfc3161Timestamps":[{"signedTimestamp":"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"}]}}, "messageSignature":{"messageDigest":{"algorithm":"SHA2_256", "digest":"bguOaDa+qFnwdRXDHvwXGfJNyqkuiiDoDflGXPlBQwg="}, "signature":"MEUCICqX1JebDFyNVpMe7pObH66zvnLMjNP3+ZU1jIlo5/RPAiEAjSUWR/hnduWQKHFTrYcA3eGrx9qehLazdrfaItMDg0k="}}

build/torch-stable-abi210-cu130-aarch64-linux/__init__.py ADDED Viewed

	@@ -0,0 +1,75 @@

+import torch
+from ._ops import ops
+from . import layers
+def silu_and_mul(out: torch.Tensor, x: torch.Tensor) -> torch.Tensor:
+    ops.silu_and_mul(out, x)
+    return out
+def mul_and_silu(out: torch.Tensor, x: torch.Tensor) -> torch.Tensor:
+    ops.mul_and_silu(out, x)
+    return out
+def gelu_and_mul(out: torch.Tensor, x: torch.Tensor) -> torch.Tensor:
+    ops.gelu_and_mul(out, x)
+    return out
+def gelu_tanh_and_mul(out: torch.Tensor, x: torch.Tensor) -> torch.Tensor:
+    ops.gelu_tanh_and_mul(out, x)
+    return out
+def fatrelu_and_mul(out: torch.Tensor, x: torch.Tensor, threshold: float = 0.0) -> torch.Tensor:
+    ops.fatrelu_and_mul(out, x, threshold)
+    return out
+def gelu(out: torch.Tensor, x: torch.Tensor) -> torch.Tensor:
+    ops.gelu(out, x)
+    return out
+def silu(out: torch.Tensor, x: torch.Tensor) -> torch.Tensor:
+    ops.silu(out, x)
+    return out
+def gelu_tanh(out: torch.Tensor, x: torch.Tensor) -> torch.Tensor:
+    ops.gelu_tanh(out, x)
+    return out
+def gelu_fast(out: torch.Tensor, x: torch.Tensor) -> torch.Tensor:
+    ops.gelu_fast(out, x)
+    return out
+def gelu_new(out: torch.Tensor, x: torch.Tensor) -> torch.Tensor:
+    ops.gelu_new(out, x)
+    return out
+def gelu_quick(out: torch.Tensor, x: torch.Tensor) -> torch.Tensor:
+    ops.gelu_quick(out, x)
+    return out
+__all__ = [
+    "silu_and_mul",
+    "mul_and_silu",
+    "gelu_and_mul",
+    "gelu_tanh_and_mul",
+    "fatrelu_and_mul",
+    "gelu_fast",
+    "gelu_new",
+    "gelu_quick",
+    "gelu_tanh",
+    "silu",
+    "gelu",
+    "layers",
+]

build/torch-stable-abi210-cu130-aarch64-linux/_activation_cuda_2f7c452.abi3.so ADDED Viewed

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

+version https://git-lfs.github.com/spec/v1
+oid sha256:dc1d1f9025d83df02c8983cd8d79c9bea97282e4d73f88034419cff2c6d9383d
+size 2662752

build/torch-stable-abi210-cu130-aarch64-linux/_ops.py ADDED Viewed

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

+import torch
+from . import _activation_cuda_2f7c452
+ops = torch.ops._activation_cuda_2f7c452
+def add_op_namespace_prefix(op_name: str):
+    """
+    Prefix op by namespace.
+    """
+    return f"_activation_cuda_2f7c452::{op_name}"

build/torch-stable-abi210-cu130-aarch64-linux/activation/__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/torch-stable-abi210-cu130-aarch64-linux/layers.py ADDED Viewed

	@@ -0,0 +1,201 @@

+import torch
+import torch.nn as nn
+from ._ops import ops
+class SiluAndMul(nn.Module):
+    """An activation function for SwiGLU.
+    The function computes x -> silu(x[:d]) * x[d:] where d = x.shape[-1] // 2.
+    Shapes:
+        x: (num_tokens, 2 * d) or (batch_size, seq_len, 2 * d)
+        return: (num_tokens, d) or (batch_size, seq_len, d)
+    """
+    can_torch_compile: bool = True
+    def forward(self, x: torch.Tensor):
+        if not x.is_contiguous():
+            x = x.contiguous()
+        d = x.shape[-1] // 2
+        output_shape = x.shape[:-1] + (d,)
+        out = torch.empty(output_shape, dtype=x.dtype, device=x.device)
+        ops.silu_and_mul(out, x)
+        return out
+class Silu(nn.Module):
+    """An activation function for SiLU.
+    The function computes x -> silu(x).
+    Shapes:
+        x: (num_tokens, d) or (batch_size, seq_len, d)
+        return: (num_tokens, d) or (batch_size, seq_len, d)
+    """
+    can_torch_compile: bool = True
+    def forward(self, x: torch.Tensor):
+        if not x.is_contiguous():
+            x = x.contiguous()
+        out = torch.empty_like(x)
+        ops.silu(out, x)
+        return out
+class Gelu(nn.Module):
+    """An activation function for GELU.
+    The function computes x -> gelu(x).
+    Shapes:
+        x: (num_tokens, d) or (batch_size, seq_len, d)
+        return: (num_tokens, d) or (batch_size, seq_len, d)
+    """
+    can_torch_compile: bool = True
+    def forward(self, x: torch.Tensor):
+        if not x.is_contiguous():
+            x = x.contiguous()
+        out = torch.empty_like(x)
+        ops.gelu(out, x)
+        return out
+class GeluTanh(nn.Module):
+    """An activation function for GELU with `tanh` approximation.
+    The function computes x -> gelu_tanh(x).
+    Shapes:
+        x: (num_tokens, d) or (batch_size, seq_len, d)
+        return: (num_tokens, d) or (batch_size, seq_len, d)
+    """
+    can_torch_compile: bool = True
+    def forward(self, x: torch.Tensor):
+        if not x.is_contiguous():
+            x = x.contiguous()
+        out = torch.empty_like(x)
+        ops.gelu_tanh(out, x)
+        return out
+class MulAndSilu(nn.Module):
+    """An activation function for SwiGLU.
+    The function computes x -> x[:d] * silu(x[d:]) where d = x.shape[-1] // 2.
+    Shapes:
+        x: (num_tokens, 2 * d) or (batch_size, seq_len, 2 * d)
+        return: (num_tokens, d) or (batch_size, seq_len, d)
+    """
+    can_torch_compile: bool = True
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        if not x.is_contiguous():
+            x = x.contiguous()
+        d = x.shape[-1] // 2
+        output_shape = x.shape[:-1] + (d,)
+        out = torch.empty(output_shape, dtype=x.dtype, device=x.device)
+        ops.mul_and_silu(out, x)
+        return out
+class GeluAndMul(nn.Module):
+    """An activation function for GeGLU.
+    The function computes x -> GELU(x[:d]) * x[d:] where d = x.shape[-1] // 2.
+    Shapes:
+        x: (batch_size, seq_len, 2 * d) or (num_tokens, 2 * d)
+        return: (batch_size, seq_len, d) or (num_tokens, d)
+    """
+    can_torch_compile: bool = True
+    def forward(self, x: torch.Tensor):
+        if not x.is_contiguous():
+            x = x.contiguous()
+        d = x.shape[-1] // 2
+        output_shape = x.shape[:-1] + (d,)
+        out = torch.empty(output_shape, dtype=x.dtype, device=x.device)
+        ops.gelu_and_mul(out, x)
+        return out
+class GeluTanhAndMul(nn.Module):
+    can_torch_compile: bool = True
+    def forward(self, x: torch.Tensor):
+        if not x.is_contiguous():
+            x = x.contiguous()
+        d = x.shape[-1] // 2
+        output_shape = x.shape[:-1] + (d,)
+        out = torch.empty(output_shape, dtype=x.dtype, device=x.device)
+        ops.gelu_tanh_and_mul(out, x)
+        return out
+class FatreluAndMul(nn.Module):
+    """An activation function for FATReLU.
+    The function computes x -> FATReLU(x[:d]) * x[d:] where
+    d = x.shape[-1] // 2.
+    This is used in openbmb/MiniCPM-S-1B-sft.
+    Shapes:
+        x: (num_tokens, 2 * d) or (batch_size, seq_len, 2 * d)
+        return: (num_tokens, d) or (batch_size, seq_len, d)
+    """
+    can_torch_compile: bool = True
+    def __init__(self, threshold: float = 0.0):
+        super().__init__()
+        self.threshold = threshold
+    def forward(self, x: torch.Tensor):
+        if not x.is_contiguous():
+            x = x.contiguous()
+        d = x.shape[-1] // 2
+        output_shape = x.shape[:-1] + (d,)
+        out = torch.empty(output_shape, dtype=x.dtype, device=x.device)
+        ops.fatrelu_and_mul(out, x, self.threshold)
+        return out
+class FastGELU(nn.Module):
+    can_torch_compile: bool = True
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        if not x.is_contiguous():
+            x = x.contiguous()
+        out = torch.empty_like(x)
+        ops.gelu_fast(out, x)
+        return out
+class NewGELU(nn.Module):
+    can_torch_compile: bool = True
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        if not x.is_contiguous():
+            x = x.contiguous()
+        out = torch.empty_like(x)
+        ops.gelu_new(out, x)
+        return out
+class QuickGELU(nn.Module):
+    can_torch_compile: bool = True
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        if not x.is_contiguous():
+            x = x.contiguous()
+        out = torch.empty_like(x)
+        ops.gelu_quick(out, x)
+        return out

build/torch-stable-abi210-cu130-aarch64-linux/metadata.json ADDED Viewed

	@@ -0,0 +1,43 @@

+{
+  "name": "activation",
+  "id": "_activation_cuda_2f7c452",
+  "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"
+    ]
+  },
+  "digest": {
+    "algorithm": "sha256",
+    "files": {
+      "__init__.py": "NQdhY7p9BPoMw0NQodMKFTPSQXBlrsX3yUdP8MEHyH0=",
+      "_activation_cuda_2f7c452.abi3.so": "3B0fkCXYPfAsiYPNjXnJvqlyguTXP4gDRBnP8sbZOD0=",
+      "_ops.py": "q9xZtwPjdsRciOCA5aQFDtsTQnV9SNcvAC74ivbShuM=",
+      "activation/__init__.py": "DFYPlrhXwYjEqCl/8n0SmWGZV8NFml5DPhMjKfv98GY=",
+      "layers.py": "wWX07+W0BNFA7RyCE0jYH7aa2U3B8aiiIpu/Q4xsE0w="
+    }
+  },
+  "provenance": {
+    "kernel-builder": {
+      "version": "0.17.0-dev0",
+      "sha": "81580bb92577f2f7228661ca2a221fc052375709",
+      "dirty": false
+    },
+    "kernel": {
+      "sha": "2f7c452801641012cf6aa9bdf042431457024033",
+      "dirty": false
+    }
+  }
+}

build/torch-stable-abi210-cu130-aarch64-linux/metadata.json.sigstore ADDED Viewed

	@@ -0,0 +1 @@

+ {"mediaType":"application/vnd.dev.sigstore.bundle.v0.3+json", "verificationMaterial":{"certificate":{"rawBytes":"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"}, "tlogEntries":[{"logIndex":"2129276940", "logId":{"keyId":"wNI9atQGlz+VWfO6LRygH4QUfY/8W4RFwiT5i5WRgB0="}, "kindVersion":{"kind":"hashedrekord", "version":"0.0.1"}, "integratedTime":"1783612187", "inclusionPromise":{"signedEntryTimestamp":"MEUCIHKxUcTk/OPDy8rzNIy60vzjSGgknoUXX9MSM+gg0n1kAiEA0D8nATRx1rNpRHoQOXfooDrrpSoIMg3xbdSdbWdEIFQ="}, "inclusionProof":{"logIndex":"2007372678", "rootHash":"PcK9X5OVNUdKxNcvVXXvF1gM809fE7mVL2/+QQuthEo=", "treeSize":"2007372680", "hashes":["N2KehzymD0KZQ8x1sySwjXE1LBYkoKOxHH0qcVRhEN0=", "H6lJGH48ZgBSSOljWzhRpg1b5AZXF/ZDa53ibyLGbwA=", "6CuFHtPU5eDGPfrnsoR9tfnGoM0FO5joaZTKo8vQYKk=", "FbAOWRmsYSIrZNgFwaGh4pN3MNumCtyxK8kozpfsp50=", "bFCHPeWT6gGx4cCjp7O8ZH7tejWmMee8KquirXRoOeI=", "slqOv5gBmQhRJrilY7eSfegXrc1VoSHecwywwt4cYoA=", "6q+LpbK/Ih183cA9KuL8+fgXqZf4ltUk/2rp524ZFuI=", "CCu7OkQW6+npAh6Tygn9JYdKku6AUboDosZP3sI7yKQ=", "uwnh4tLrO/3+ywQ63CnFW6mU2V+y+Db8CobG3g6S0Hk=", "Kezfh1XcoOp9KAwr7OWyF8C7cr/mZBPGqKHTdO0BnDs=", "LD5Ws0n7JWm8O7qGvHNSZug32ByFM1WblCpsDJh5M6o=", "f8knrc4LOu/X8scT5SBvEgFJSd7vZZ0LzSaC0Z3lm/U=", "WajL3ZhY6ZX9meV3VTH9TkrbM31OF4o6+7foL0Uf6PI=", "cyKMKBvHkQtwb2xR5svh+YppANOVdn7X8t4DuKMxICA=", "+/VZ56MsIPxMiyLAodzKXo5TEWdQp36z89qLhpzloAo=", "daxmZaajRpZV+JxHiOYZhJBiSKN5ucqjh2WnGbHhirw=", "DOCeoSMovIvLExkhIvisow9AuNXgeWs4ECkyR6EcqYU="], "checkpoint":{"envelope":"rekor.sigstore.dev - 1193050959916656506\n2007372680\nPcK9X5OVNUdKxNcvVXXvF1gM809fE7mVL2/+QQuthEo=\n\n— rekor.sigstore.dev wNI9ajBEAiADQ2GfI43pB2yAyAml79Upy92/nKzQgnv6pvKABpIpKAIgdFGqDVisnJNWTnkXy/L6D6XGGkV4LQVczSRv2/cdsKI=\n"}}, "canonicalizedBody":"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"}], "timestampVerificationData":{"rfc3161Timestamps":[{"signedTimestamp":"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"}]}}, "messageSignature":{"messageDigest":{"algorithm":"SHA2_256", "digest":"+UrS8sQH6z1/aV6DDNyqvLSEoZUh4pGAGiErXaHhR4M="}, "signature":"MEQCIA/jFId8JX4o+bjo6jMz91ROQHzTqw6R35wHPnpD7IHkAiAG7UYsGGOCwQaFQoO78Phdtz51ZQA9k50pBd8YI8YzlQ=="}}

build/torch-stable-abi210-cu132-aarch64-linux/__init__.py ADDED Viewed

	@@ -0,0 +1,75 @@

+import torch
+from ._ops import ops
+from . import layers
+def silu_and_mul(out: torch.Tensor, x: torch.Tensor) -> torch.Tensor:
+    ops.silu_and_mul(out, x)
+    return out
+def mul_and_silu(out: torch.Tensor, x: torch.Tensor) -> torch.Tensor:
+    ops.mul_and_silu(out, x)
+    return out
+def gelu_and_mul(out: torch.Tensor, x: torch.Tensor) -> torch.Tensor:
+    ops.gelu_and_mul(out, x)
+    return out
+def gelu_tanh_and_mul(out: torch.Tensor, x: torch.Tensor) -> torch.Tensor:
+    ops.gelu_tanh_and_mul(out, x)
+    return out
+def fatrelu_and_mul(out: torch.Tensor, x: torch.Tensor, threshold: float = 0.0) -> torch.Tensor:
+    ops.fatrelu_and_mul(out, x, threshold)
+    return out
+def gelu(out: torch.Tensor, x: torch.Tensor) -> torch.Tensor:
+    ops.gelu(out, x)
+    return out
+def silu(out: torch.Tensor, x: torch.Tensor) -> torch.Tensor:
+    ops.silu(out, x)
+    return out
+def gelu_tanh(out: torch.Tensor, x: torch.Tensor) -> torch.Tensor:
+    ops.gelu_tanh(out, x)
+    return out
+def gelu_fast(out: torch.Tensor, x: torch.Tensor) -> torch.Tensor:
+    ops.gelu_fast(out, x)
+    return out
+def gelu_new(out: torch.Tensor, x: torch.Tensor) -> torch.Tensor:
+    ops.gelu_new(out, x)
+    return out
+def gelu_quick(out: torch.Tensor, x: torch.Tensor) -> torch.Tensor:
+    ops.gelu_quick(out, x)
+    return out
+__all__ = [
+    "silu_and_mul",
+    "mul_and_silu",
+    "gelu_and_mul",
+    "gelu_tanh_and_mul",
+    "fatrelu_and_mul",
+    "gelu_fast",
+    "gelu_new",
+    "gelu_quick",
+    "gelu_tanh",
+    "silu",
+    "gelu",
+    "layers",
+]

build/torch-stable-abi210-cu132-aarch64-linux/_activation_cuda_2f7c452.abi3.so ADDED Viewed

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

+version https://git-lfs.github.com/spec/v1
+oid sha256:2e3af076e8fc0c050f4635d903f191310808c92345fd1289c95ae77c54076f10
+size 2730976

build/torch-stable-abi210-cu132-aarch64-linux/_ops.py ADDED Viewed

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

+import torch
+from . import _activation_cuda_2f7c452
+ops = torch.ops._activation_cuda_2f7c452
+def add_op_namespace_prefix(op_name: str):
+    """
+    Prefix op by namespace.
+    """
+    return f"_activation_cuda_2f7c452::{op_name}"

build/torch-stable-abi210-cu132-aarch64-linux/activation/__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/torch-stable-abi210-cu132-aarch64-linux/layers.py ADDED Viewed

	@@ -0,0 +1,201 @@

+import torch
+import torch.nn as nn
+from ._ops import ops
+class SiluAndMul(nn.Module):
+    """An activation function for SwiGLU.
+    The function computes x -> silu(x[:d]) * x[d:] where d = x.shape[-1] // 2.
+    Shapes:
+        x: (num_tokens, 2 * d) or (batch_size, seq_len, 2 * d)
+        return: (num_tokens, d) or (batch_size, seq_len, d)
+    """
+    can_torch_compile: bool = True
+    def forward(self, x: torch.Tensor):
+        if not x.is_contiguous():
+            x = x.contiguous()
+        d = x.shape[-1] // 2
+        output_shape = x.shape[:-1] + (d,)
+        out = torch.empty(output_shape, dtype=x.dtype, device=x.device)
+        ops.silu_and_mul(out, x)
+        return out
+class Silu(nn.Module):
+    """An activation function for SiLU.
+    The function computes x -> silu(x).
+    Shapes:
+        x: (num_tokens, d) or (batch_size, seq_len, d)
+        return: (num_tokens, d) or (batch_size, seq_len, d)
+    """
+    can_torch_compile: bool = True
+    def forward(self, x: torch.Tensor):
+        if not x.is_contiguous():
+            x = x.contiguous()
+        out = torch.empty_like(x)
+        ops.silu(out, x)
+        return out
+class Gelu(nn.Module):
+    """An activation function for GELU.
+    The function computes x -> gelu(x).
+    Shapes:
+        x: (num_tokens, d) or (batch_size, seq_len, d)
+        return: (num_tokens, d) or (batch_size, seq_len, d)
+    """
+    can_torch_compile: bool = True
+    def forward(self, x: torch.Tensor):
+        if not x.is_contiguous():
+            x = x.contiguous()
+        out = torch.empty_like(x)
+        ops.gelu(out, x)
+        return out
+class GeluTanh(nn.Module):
+    """An activation function for GELU with `tanh` approximation.
+    The function computes x -> gelu_tanh(x).
+    Shapes:
+        x: (num_tokens, d) or (batch_size, seq_len, d)
+        return: (num_tokens, d) or (batch_size, seq_len, d)
+    """
+    can_torch_compile: bool = True
+    def forward(self, x: torch.Tensor):
+        if not x.is_contiguous():
+            x = x.contiguous()
+        out = torch.empty_like(x)
+        ops.gelu_tanh(out, x)
+        return out
+class MulAndSilu(nn.Module):
+    """An activation function for SwiGLU.
+    The function computes x -> x[:d] * silu(x[d:]) where d = x.shape[-1] // 2.
+    Shapes:
+        x: (num_tokens, 2 * d) or (batch_size, seq_len, 2 * d)
+        return: (num_tokens, d) or (batch_size, seq_len, d)
+    """
+    can_torch_compile: bool = True
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        if not x.is_contiguous():
+            x = x.contiguous()
+        d = x.shape[-1] // 2
+        output_shape = x.shape[:-1] + (d,)
+        out = torch.empty(output_shape, dtype=x.dtype, device=x.device)
+        ops.mul_and_silu(out, x)
+        return out
+class GeluAndMul(nn.Module):
+    """An activation function for GeGLU.
+    The function computes x -> GELU(x[:d]) * x[d:] where d = x.shape[-1] // 2.
+    Shapes:
+        x: (batch_size, seq_len, 2 * d) or (num_tokens, 2 * d)
+        return: (batch_size, seq_len, d) or (num_tokens, d)
+    """
+    can_torch_compile: bool = True
+    def forward(self, x: torch.Tensor):
+        if not x.is_contiguous():
+            x = x.contiguous()
+        d = x.shape[-1] // 2
+        output_shape = x.shape[:-1] + (d,)
+        out = torch.empty(output_shape, dtype=x.dtype, device=x.device)
+        ops.gelu_and_mul(out, x)
+        return out
+class GeluTanhAndMul(nn.Module):
+    can_torch_compile: bool = True
+    def forward(self, x: torch.Tensor):
+        if not x.is_contiguous():
+            x = x.contiguous()
+        d = x.shape[-1] // 2
+        output_shape = x.shape[:-1] + (d,)
+        out = torch.empty(output_shape, dtype=x.dtype, device=x.device)
+        ops.gelu_tanh_and_mul(out, x)
+        return out
+class FatreluAndMul(nn.Module):
+    """An activation function for FATReLU.
+    The function computes x -> FATReLU(x[:d]) * x[d:] where
+    d = x.shape[-1] // 2.
+    This is used in openbmb/MiniCPM-S-1B-sft.
+    Shapes:
+        x: (num_tokens, 2 * d) or (batch_size, seq_len, 2 * d)
+        return: (num_tokens, d) or (batch_size, seq_len, d)
+    """
+    can_torch_compile: bool = True
+    def __init__(self, threshold: float = 0.0):
+        super().__init__()
+        self.threshold = threshold
+    def forward(self, x: torch.Tensor):
+        if not x.is_contiguous():
+            x = x.contiguous()
+        d = x.shape[-1] // 2
+        output_shape = x.shape[:-1] + (d,)
+        out = torch.empty(output_shape, dtype=x.dtype, device=x.device)
+        ops.fatrelu_and_mul(out, x, self.threshold)
+        return out
+class FastGELU(nn.Module):
+    can_torch_compile: bool = True
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        if not x.is_contiguous():
+            x = x.contiguous()
+        out = torch.empty_like(x)
+        ops.gelu_fast(out, x)
+        return out
+class NewGELU(nn.Module):
+    can_torch_compile: bool = True
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        if not x.is_contiguous():
+            x = x.contiguous()
+        out = torch.empty_like(x)
+        ops.gelu_new(out, x)
+        return out
+class QuickGELU(nn.Module):
+    can_torch_compile: bool = True
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        if not x.is_contiguous():
+            x = x.contiguous()
+        out = torch.empty_like(x)
+        ops.gelu_quick(out, x)
+        return out

build/torch-stable-abi210-cu132-aarch64-linux/metadata.json ADDED Viewed

	@@ -0,0 +1,43 @@

+{
+  "name": "activation",
+  "id": "_activation_cuda_2f7c452",
+  "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"
+    ]
+  },
+  "digest": {
+    "algorithm": "sha256",
+    "files": {
+      "__init__.py": "NQdhY7p9BPoMw0NQodMKFTPSQXBlrsX3yUdP8MEHyH0=",
+      "_activation_cuda_2f7c452.abi3.so": "Ljrwduj8DAUPRjXZA/GRMQgIySNF/RKJyVrnfFQHbxA=",
+      "_ops.py": "q9xZtwPjdsRciOCA5aQFDtsTQnV9SNcvAC74ivbShuM=",
+      "activation/__init__.py": "DFYPlrhXwYjEqCl/8n0SmWGZV8NFml5DPhMjKfv98GY=",
+      "layers.py": "wWX07+W0BNFA7RyCE0jYH7aa2U3B8aiiIpu/Q4xsE0w="
+    }
+  },
+  "provenance": {
+    "kernel-builder": {
+      "version": "0.17.0-dev0",
+      "sha": "81580bb92577f2f7228661ca2a221fc052375709",
+      "dirty": false
+    },
+    "kernel": {
+      "sha": "2f7c452801641012cf6aa9bdf042431457024033",
+      "dirty": false
+    }
+  }
+}

build/torch-stable-abi210-cu132-aarch64-linux/metadata.json.sigstore ADDED Viewed

	@@ -0,0 +1 @@

+ {"mediaType":"application/vnd.dev.sigstore.bundle.v0.3+json", "verificationMaterial":{"certificate":{"rawBytes":"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"}, "tlogEntries":[{"logIndex":"2129277158", "logId":{"keyId":"wNI9atQGlz+VWfO6LRygH4QUfY/8W4RFwiT5i5WRgB0="}, "kindVersion":{"kind":"hashedrekord", "version":"0.0.1"}, "integratedTime":"1783612188", "inclusionPromise":{"signedEntryTimestamp":"MEYCIQC+Dimw0xr66PFvq6eXMSNm/O8h52LHvNhtsE7ap9SJYgIhAOMcfxHGYFTEWa2LQZqGyjBo3juDSXJS2NZUpRNG1DkH"}, "inclusionProof":{"logIndex":"2007372896", "rootHash":"LV+qAY9J0YWMwdcQAF0BnWpbJcoVgZNNt0lEHNgLK94=", "treeSize":"2007372928", "hashes":["kN6iJbTtEZvjkSojx5K1EkLxWVoQzWug1B/o7WcUJ68=", "DU81dCIR3NZ+rjEltK9mhT8oTH0LEltmoM3xuFE0HjY=", "h32s6rPhHO9lD6aha+7D3ZWfn4DJuR9I98QvuIEOtFA=", "H0VvEBumWCVY6FMTNVCiNH0Jng1ihn6MTdI347gMWM8=", "ER5iBhpNxBgIO8dh9lrLy/9hkFwjtWzss9GYA9NnwPM=", "6W8vJ+D6fTDBxgGB2l2QeKI23npi+2klImqRgQoaj/8=", "bizkQtow9bYDIzaYrKj7dS5AYaxQiswPSfH+0TnuQ88=", "59rSn8Rt0cVWqgReMM0Yi86oeneXlh+FZXICIaZzV0Y=", "6q+LpbK/Ih183cA9KuL8+fgXqZf4ltUk/2rp524ZFuI=", "CCu7OkQW6+npAh6Tygn9JYdKku6AUboDosZP3sI7yKQ=", "uwnh4tLrO/3+ywQ63CnFW6mU2V+y+Db8CobG3g6S0Hk=", "Kezfh1XcoOp9KAwr7OWyF8C7cr/mZBPGqKHTdO0BnDs=", "LD5Ws0n7JWm8O7qGvHNSZug32ByFM1WblCpsDJh5M6o=", "f8knrc4LOu/X8scT5SBvEgFJSd7vZZ0LzSaC0Z3lm/U=", "WajL3ZhY6ZX9meV3VTH9TkrbM31OF4o6+7foL0Uf6PI=", "cyKMKBvHkQtwb2xR5svh+YppANOVdn7X8t4DuKMxICA=", "+/VZ56MsIPxMiyLAodzKXo5TEWdQp36z89qLhpzloAo=", "daxmZaajRpZV+JxHiOYZhJBiSKN5ucqjh2WnGbHhirw=", "DOCeoSMovIvLExkhIvisow9AuNXgeWs4ECkyR6EcqYU="], "checkpoint":{"envelope":"rekor.sigstore.dev - 1193050959916656506\n2007372928\nLV+qAY9J0YWMwdcQAF0BnWpbJcoVgZNNt0lEHNgLK94=\n\n— rekor.sigstore.dev wNI9ajBFAiEA33x0fRKyBoqxawtiG3WTjzhl+rVFYO2v5CfJzoqyFdgCIGQkujp6KwvHrKRm0g80hwNEDRM1unM+39PHS4cghPJt\n"}}, "canonicalizedBody":"eyJhcGlWZXJzaW9uIjoiMC4wLjEiLCJraW5kIjoiaGFzaGVkcmVrb3JkIiwic3BlYyI6eyJkYXRhIjp7Imhhc2giOnsiYWxnb3JpdGhtIjoic2hhMjU2IiwidmFsdWUiOiJlMDgyNmYxZGE3NTVkZmUyOTk4MzVmYzlkMTUxOWZmMzU5NWJmMWY2YmU4NWY5ZmFlNjYzNGQxNGU5ZTdhMWFhIn19LCJzaWduYXR1cmUiOnsiY29udGVudCI6Ik1FUUNJRGJNZ2dJWjEyN012d1ZXa0FQNVN0ek15TGwrVVhxc0NMRXBXVmh4OGw5bkFpQmhKdTUvaTlvUjlDWXdYVytrS2JFMG01RVh4QlRpRHJaTE5haC9NTmFnNWc9PSIsInB1YmxpY0tleSI6eyJjb250ZW50IjoiTFMwdExTMUNSVWRKVGlCRFJWSlVTVVpKUTBGVVJTMHRMUzB0Q2sxSlNVaFRla05EUW5SSFowRjNTVUpCWjBsVlpreDRTVFJIWkdOVVdFTnhWMlZaUVRrMlpISTJTMmxMZVRjd2QwTm5XVWxMYjFwSmVtb3dSVUYzVFhjS1RucEZWazFDVFVkQk1WVkZRMmhOVFdNeWJHNWpNMUoyWTIxVmRWcEhWakpOVWpSM1NFRlpSRlpSVVVSRmVGWjZZVmRrZW1SSE9YbGFVekZ3WW01U2JBcGpiVEZzV2tkc2FHUkhWWGRJYUdOT1RXcFpkMDU2UVRWTlZGVXdUMVJSTTFkb1kwNU5hbGwzVG5wQk5VMVVWVEZQVkZFelYycEJRVTFHYTNkRmQxbElDa3R2V2tsNmFqQkRRVkZaU1V0dldrbDZhakJFUVZGalJGRm5RVVZVY1VKeFozWlpNMUpLUW04emJGRkVSelJTYm1wc1JrTmhlWGRKUjJ4WVpsTlJiVWdLVFRSTlpqUnZWVGxpTkZvemVIaFROMG8wV1ZkclVHcGpVVWR3YVZGbllUUnVabGN4TmxKVFVHNVFlVVJOVFhodVFrdFBRMEptUVhkbloxaHpUVUUwUndwQk1WVmtSSGRGUWk5M1VVVkJkMGxJWjBSQlZFSm5UbFpJVTFWRlJFUkJTMEpuWjNKQ1owVkdRbEZqUkVGNlFXUkNaMDVXU0ZFMFJVWm5VVlZCV1RSR0NrVXdNblU1WVM5QmQySmtVWGRSYlcwMWFDdE1TRms0ZDBoM1dVUldVakJxUWtKbmQwWnZRVlV6T1ZCd2VqRlphMFZhWWpWeFRtcHdTMFpYYVhocE5Ga0tXa1E0ZDJGM1dVUldVakJTUVZGSUwwSkhSWGRZTkZwa1lVaFNNR05JVFRaTWVUbHVZVmhTYjJSWFNYVlpNamwwVERKb01Wb3laSEJpYldSdFdWZE9iQXBNTW5Sc1kyMDFiR0pJVFhSWk1qbDBZbGhXZFdGWVVqVk1lVFZ1WVZoU2IyUlhTWFprTWpsNVlUSmFjMkl6WkhwTU1rb3hZVmQ0YTB4dWJHaGlWM2hCQ21OdFZtMWplVGx2V2xkR2EyTjVPWFJaVjJ4MVRVUnJSME5wYzBkQlVWRkNaemM0ZDBGUlJVVkxNbWd3WkVoQ2VrOXBPSFprUnpseVdsYzBkVmxYVGpBS1lWYzVkV041Tlc1aFdGSnZaRmRLTVdNeVZubFpNamwxWkVkV2RXUkROV3BpTWpCM1NIZFpTMHQzV1VKQ1FVZEVkbnBCUWtGblVWSmtNamw1WVRKYWN3cGlNMlJtV2tkc2VtTkhSakJaTW1kM1RtZFpTMHQzV1VKQ1FVZEVkbnBCUWtGM1VXOU5iVmt6V1hwUk1VMXFaM2ROVkZrd1RWUkJlRTF0VG0xT2JVWm9DazlYU210YWFrRXdUV3BSZWsxVVVURk9la0Y1VGtSQmVrMTZRVlJDWjI5eVFtZEZSVUZaVHk5TlFVVkZRa0ZXUTJSWGJITmFSRUZ5UW1kdmNrSm5SVVVLUVZsUEwwMUJSVVpDUWpGdlpGZGtibUZYTlc1YWJVWnFXbE01Y2xwWVNuVmFWM2g2VEZkT2RtSlhNVEZpYld3d1pWUkJaRUpuYjNKQ1owVkZRVmxQTHdwTlFVVkhRa0U1ZVZwWFducE1NbWhzV1ZkU2Vrd3lNV2hoVnpSM1QzZFpTMHQzV1VKQ1FVZEVkbnBCUWtOQlVYUkVRM1J2WkVoU2QyTjZiM1pNTTFKMkNtRXlWblZNYlVacVpFZHNkbUp1VFhWYU1td3dZVWhXYVdSWVRteGpiVTUyWW01U2JHSnVVWFZaTWpsMFRVY3dSME5wYzBkQlVWRkNaemM0ZDBGUmEwVUtXSGQ0WkdGSVVqQmpTRTAyVEhrNWJtRllVbTlrVjBsMVdUSTVkRXd5YURGYU1tUndZbTFrYlZsWFRteE1NblJzWTIwMWJHSklUWFJaTWpsMFlsaFdkUXBoV0ZJMVRIazFibUZZVW05a1YwbDJaREk1ZVdFeVduTmlNMlI2VERKS01XRlhlR3RNYm14b1lsZDRRV050Vm0xamVUbHZXbGRHYTJONU9YUlpWMngxQ2sxRVowZERhWE5IUVZGUlFtYzNPSGRCVVc5RlMyZDNiMDF0V1ROWmVsRXhUV3BuZDAxVVdUQk5WRUY0VFcxT2JVNXRSbWhQVjBwcldtcEJNRTFxVVhvS1RWUlJNVTU2UVhsT1JFRjZUWHBCWWtKbmIzSkNaMFZGUVZsUEwwMUJSVXhDUVRCTlF6Tk9iR0pIV1hSaFJ6bDZaRWRXYTAxRlFVZERhWE5IUVZGUlFncG5OemgzUVZGM1JVMW5kM2RoU0ZJd1kwaE5Oa3g1T1c1aFdGSnZaRmRKZFZreU9YUk1NbWd4V2pKa2NHSnRaRzFaVjA1c1RESjBiR050Tld4aVNFMTBDbGt5T1hSaVdGWjFZVmhTTlUxRVowZERhWE5IUVZGUlFtYzNPSGRCVVRCRlMyZDNiMDF0V1ROWmVsRXhUV3BuZDAxVVdUQk5WRUY0VFcxT2JVNXRSbWdLVDFkS2ExcHFRVEJOYWxGNlRWUlJNVTU2UVhsT1JFRjZUWHBCWmtKbmIzSkNaMFZGUVZsUEwwMUJSVTlDUWtWTlJETktiRnB1VFhaaFIxWm9Xa2hOZGdwaVYwWndZbXBCWVVKbmIzSkNaMFZGUVZsUEwwMUJSVkJDUVhkTlEycEZkMDU2UlRCT2VsVXhUV3ByZDB4bldVdExkMWxDUWtGSFJIWjZRVUpGUVZGbkNrUkNOVzlrU0ZKM1kzcHZka3d5WkhCa1IyZ3hXV2sxYW1JeU1IWmhTRlp1V2pKc2RWb3lXbWhaTWxWM1IwRlpTMHQzV1VKQ1FVZEVkbnBCUWtWUlVVc0tSRUZuZVU1VVkzbE5SR013VFhwQ2RFSm5iM0pDWjBWRlFWbFBMMDFCUlZOQ1JqaE5XRmRvTUdSSVFucFBhVGgyV2pKc01HRklWbWxNYlU1MllsTTVid3BrVjJSdVlWYzFibHB0Um1wYVV6bHlXbGhLZFZwWGVIcE1WMDUyWWxjeE1XSnRiREJsVXpoMVdqSnNNR0ZJVm1sTU0yUjJZMjEwYldKSE9UTmplVGxwQ21SWGJITmFRelUxV1ZjeGMxRklTbXhhYmsxMllVZFdhRnBJVFhaaVYwWndZbXBCTkVKbmIzSkNaMFZGUVZsUEwwMUJSVlJDUTI5TlMwUktiVTR5VFRBS1RsUkpORTFFUlRKT1JFVjNUVlJLYWxwcVdtaFpWR3hwV2tkWmQwNUVTVEJOZWtVd1RsUmpkMDFxVVhkTmVrMTNTVkZaUzB0M1dVSkNRVWRFZG5wQlFncEdRVkZVUkVKR00ySXpTbkphYlhoMlpERTVhMkZZVG5kWldGSnFZVVJDYTBKbmIzSkNaMFZGUVZsUEwwMUJSVlpDUmxsTlZrZG9NR1JJUW5wUGFUaDJDbG95YkRCaFNGWnBURzFPZG1KVE9XOWtWMlJ1WVZjMWJscHRSbXBhVXpseVdsaEtkVnBYZUhwTVYwNTJZbGN4TVdKdGJEQmxVemxvV1ROU2NHSXlOWG9LVEROS01XSnVUWFpOYW10M1RYcEJNMDU2UlhoTlJFMTJXVmhTTUZwWE1YZGtTRTEyVFZSQlYwSm5iM0pDWjBWRlFWbFBMMDFCUlZkQ1FXZE5RbTVDTVFwWmJYaHdXWHBDUjBKbmIzSkNaMFZGUVZsUEwwMUJSVmxDUkdkTlRtNUtiR05IT0RaaFNGWnVXakpzZFZveVdtaFpNbFYyWVRKV2VXSnRWbk5qZVRGcUNtSXlNWFJrVnpWd1pFaHJObU50Vm0xUGJrcHNXbTVOZG1GSFZtaGFTRTEyWWxkR2NHSnFRMEpwWjFsTFMzZFpRa0pCU0ZkbFVVbEZRV2RTT0VKSWIwRUtaVUZDTWtGT01EbE5SM0pIZUhoRmVWbDRhMlZJU214dVRuZExhVk5zTmpRemFubDBMelJsUzJOdlFYWkxaVFpQUVVGQlFtNHdaVk5LVTJOQlFVRlJSQXBCUldOM1VsRkphRUZMVXpKd01uRlhVMlU1SzJSeU9UVllUVUkyZUU1VmNVMUpWV2xzVEdORk5WQmxLMDFrSzA1d2IyUkxRV2xCUzIxbVZYTkZXVlJhQ25KS2J5OW1hRzFZUldsYWVtaFFWRTB5TVZoNGQySXlTemhXVmt4ck9GQkxOa1JCUzBKblozRm9hMnBQVUZGUlJFRjNUbTlCUkVKc1FXcEZRWGxXV2pjS1FYUkpjWEUzYTNaYWNVc3dOV0Z3YWxKelZHRTFkMVYxTnpSbGFVOHpZV051ZUhWTWVXRk5VMHhITDFNek9WazRkRVJrVTJGeU1EQjNjWFI1UVdwQmR3cE9jVVJhY21oUWJtNXVRMFZvVFZSa1EwOXhWVWxxVFU1YVNWQlBSMFJ1Y2tkS1RHRnlVVWwzTXpRMVMwdG5ObFpTTTJNM1VURkhWM055UVhwRFozYzlDaTB0TFMwdFJVNUVJRU5GVWxSSlJrbERRVlJGTFMwdExTMEsifX19fQ=="}], "timestampVerificationData":{"rfc3161Timestamps":[{"signedTimestamp":"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"}]}}, "messageSignature":{"messageDigest":{"algorithm":"SHA2_256", "digest":"4IJvHadV3+KZg1/J0VGf81lb8fa+hfn65mNNFOnnoao="}, "signature":"MEQCIDbMggIZ127MvwVWkAP5StzMyLl+UXqsCLEpWVhx8l9nAiBhJu5/i9oR9CYwXW+kKbE0m5EXxBTiDrZLNah/MNag5g=="}}