add build

build/torch27-cxx11-cu118-x86_64-linux/activation/{_activation_cf68df1_dirty.abi3.so → _activation_f517c97_dirty.abi3.so}

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1655e52503ce7d0b7dabd55b97c1bd7d11071cbe0f80b9e810c443523638fd9b
-size 2994312
+oid sha256:bd84c828d4c15e96d65d6c8f0eb7a945ee8167d92e978b2ebce03eeaf41e7fce
+size 4405112

build/torch27-cxx11-cu118-x86_64-linux/activation/_ops.py

@@ -1,9 +1,9 @@
 import torch
-from . import _activation_cf68df1_dirty
-ops = torch.ops._activation_cf68df1_dirty
+from . import _activation_f517c97_dirty
+ops = torch.ops._activation_f517c97_dirty
 
 def add_op_namespace_prefix(op_name: str):
     """
     Prefix op by namespace.
     """
-    return f"_activation_cf68df1_dirty::{op_name}"
+    return f"_activation_f517c97_dirty::{op_name}"

build/torch27-cxx11-cu118-x86_64-linux/activation/layers.py

@@ -7,6 +7,7 @@ from .rms_norm import RMSNormFunction
 
 
 class PolyNorm(nn.Module):
+
     def __init__(self, eps=1e-6, dtype: torch.dtype = torch.float32):
         super().__init__()
         self.weight = torch.nn.Parameter(torch.ones(3, dtype=dtype) / 3)
@@ -28,6 +29,7 @@ class PolyNorm(nn.Module):
 
 
 class RMSNorm(nn.Module):
+
     def __init__(self, dim: int, eps=1e-6, dtype: torch.dtype = torch.float32):
         super().__init__()
         self.weight = torch.nn.Parameter(torch.ones(dim, dtype=dtype))

build/torch27-cxx11-cu118-x86_64-linux/activation/poly_norm.py

@@ -26,16 +26,14 @@ class PolyNormFunction(torch.autograd.Function):
         input, weight = ctx.saved_tensors
         eps = ctx.eps
 
-        input_grad = torch.empty_like(input) if ctx.needs_input_grad[0] else None
-        weight_grad = torch.empty_like(weight) if ctx.needs_input_grad[1] else None
-        bias_grad = (
-            torch.empty(1, dtype=weight.dtype, device=weight.device)
-            if ctx.needs_input_grad[2]
-            else None
-        )
-
-        ops.poly_norm_backward(
-            input_grad, weight_grad, bias_grad, output_grad, input, weight, eps
-        )
+        input_grad = torch.empty_like(
+            input) if ctx.needs_input_grad[0] else None
+        weight_grad = torch.empty_like(
+            weight) if ctx.needs_input_grad[1] else None
+        bias_grad = (torch.empty(1, dtype=weight.dtype, device=weight.device)
+                     if ctx.needs_input_grad[2] else None)
+
+        ops.poly_norm_backward(input_grad, weight_grad, bias_grad, output_grad,
+                               input, weight, eps)
 
         return input_grad, weight_grad, bias_grad, None

build/torch27-cxx11-cu118-x86_64-linux/activation/rms_norm.py

@@ -26,9 +26,12 @@ class RMSNormFunction(torch.autograd.Function):
         input, weight = ctx.saved_tensors
         eps = ctx.eps
 
-        input_grad = torch.empty_like(input) if ctx.needs_input_grad[0] else None
-        weight_grad = torch.empty_like(weight) if ctx.needs_input_grad[1] else None
+        input_grad = torch.empty_like(
+            input) if ctx.needs_input_grad[0] else None
+        weight_grad = torch.empty_like(
+            weight) if ctx.needs_input_grad[1] else None
 
-        ops.rms_norm_backward(input_grad, weight_grad, output_grad, input, weight, eps)
+        ops.rms_norm_backward(input_grad, weight_grad, output_grad, input,
+                              weight, eps)
 
         return input_grad, weight_grad, None

build/torch27-cxx11-cu126-x86_64-linux/activation/{_activation_cf68df1_dirty.abi3.so → _activation_f517c97_dirty.abi3.so}

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:174dbe4375aa22fb34d9d23630b3bec4eeb95635ef681b665db0985e78cf5af3
-size 3027504
+oid sha256:caffcadbb99fbaa27e8a81d5ef508f2e1a798e7626d618c3cf5b0d387d2c8686
+size 4618624

build/torch27-cxx11-cu126-x86_64-linux/activation/_ops.py

@@ -1,9 +1,9 @@
 import torch
-from . import _activation_cf68df1_dirty
-ops = torch.ops._activation_cf68df1_dirty
+from . import _activation_f517c97_dirty
+ops = torch.ops._activation_f517c97_dirty
 
 def add_op_namespace_prefix(op_name: str):
     """
     Prefix op by namespace.
     """
-    return f"_activation_cf68df1_dirty::{op_name}"
+    return f"_activation_f517c97_dirty::{op_name}"

build/torch27-cxx11-cu126-x86_64-linux/activation/layers.py

@@ -7,6 +7,7 @@ from .rms_norm import RMSNormFunction
 
 
 class PolyNorm(nn.Module):
+
     def __init__(self, eps=1e-6, dtype: torch.dtype = torch.float32):
         super().__init__()
         self.weight = torch.nn.Parameter(torch.ones(3, dtype=dtype) / 3)
@@ -28,6 +29,7 @@ class PolyNorm(nn.Module):
 
 
 class RMSNorm(nn.Module):
+
     def __init__(self, dim: int, eps=1e-6, dtype: torch.dtype = torch.float32):
         super().__init__()
         self.weight = torch.nn.Parameter(torch.ones(dim, dtype=dtype))

build/torch27-cxx11-cu126-x86_64-linux/activation/poly_norm.py

@@ -26,16 +26,14 @@ class PolyNormFunction(torch.autograd.Function):
         input, weight = ctx.saved_tensors
         eps = ctx.eps
 
-        input_grad = torch.empty_like(input) if ctx.needs_input_grad[0] else None
-        weight_grad = torch.empty_like(weight) if ctx.needs_input_grad[1] else None
-        bias_grad = (
-            torch.empty(1, dtype=weight.dtype, device=weight.device)
-            if ctx.needs_input_grad[2]
-            else None
-        )
-
-        ops.poly_norm_backward(
-            input_grad, weight_grad, bias_grad, output_grad, input, weight, eps
-        )
+        input_grad = torch.empty_like(
+            input) if ctx.needs_input_grad[0] else None
+        weight_grad = torch.empty_like(
+            weight) if ctx.needs_input_grad[1] else None
+        bias_grad = (torch.empty(1, dtype=weight.dtype, device=weight.device)
+                     if ctx.needs_input_grad[2] else None)
+
+        ops.poly_norm_backward(input_grad, weight_grad, bias_grad, output_grad,
+                               input, weight, eps)
 
         return input_grad, weight_grad, bias_grad, None

build/torch27-cxx11-cu126-x86_64-linux/activation/rms_norm.py

@@ -26,9 +26,12 @@ class RMSNormFunction(torch.autograd.Function):
         input, weight = ctx.saved_tensors
         eps = ctx.eps
 
-        input_grad = torch.empty_like(input) if ctx.needs_input_grad[0] else None
-        weight_grad = torch.empty_like(weight) if ctx.needs_input_grad[1] else None
+        input_grad = torch.empty_like(
+            input) if ctx.needs_input_grad[0] else None
+        weight_grad = torch.empty_like(
+            weight) if ctx.needs_input_grad[1] else None
 
-        ops.rms_norm_backward(input_grad, weight_grad, output_grad, input, weight, eps)
+        ops.rms_norm_backward(input_grad, weight_grad, output_grad, input,
+                              weight, eps)
 
         return input_grad, weight_grad, None

build/torch27-cxx11-cu128-x86_64-linux/activation/{_activation_cf68df1_dirty.abi3.so → _activation_f517c97_dirty.abi3.so}

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:91d71ca84a19b393c22b269226a7b4ddadbf1feec73a80bd45f655179c7a53f5
-size 3987512
+oid sha256:3b7c6ece8e8d316c4cc5fe46b1cec4422b2f61e9bb7240af71a2b4a35975d8e6
+size 6676528

build/torch27-cxx11-cu128-x86_64-linux/activation/_ops.py

@@ -1,9 +1,9 @@
 import torch
-from . import _activation_cf68df1_dirty
-ops = torch.ops._activation_cf68df1_dirty
+from . import _activation_f517c97_dirty
+ops = torch.ops._activation_f517c97_dirty
 
 def add_op_namespace_prefix(op_name: str):
     """
     Prefix op by namespace.
     """
-    return f"_activation_cf68df1_dirty::{op_name}"
+    return f"_activation_f517c97_dirty::{op_name}"

build/torch27-cxx11-cu128-x86_64-linux/activation/layers.py

@@ -7,6 +7,7 @@ from .rms_norm import RMSNormFunction
 
 
 class PolyNorm(nn.Module):
+
     def __init__(self, eps=1e-6, dtype: torch.dtype = torch.float32):
         super().__init__()
         self.weight = torch.nn.Parameter(torch.ones(3, dtype=dtype) / 3)
@@ -28,6 +29,7 @@ class PolyNorm(nn.Module):
 
 
 class RMSNorm(nn.Module):
+
     def __init__(self, dim: int, eps=1e-6, dtype: torch.dtype = torch.float32):
         super().__init__()
         self.weight = torch.nn.Parameter(torch.ones(dim, dtype=dtype))

build/torch27-cxx11-cu128-x86_64-linux/activation/poly_norm.py

@@ -26,16 +26,14 @@ class PolyNormFunction(torch.autograd.Function):
         input, weight = ctx.saved_tensors
         eps = ctx.eps
 
-        input_grad = torch.empty_like(input) if ctx.needs_input_grad[0] else None
-        weight_grad = torch.empty_like(weight) if ctx.needs_input_grad[1] else None
-        bias_grad = (
-            torch.empty(1, dtype=weight.dtype, device=weight.device)
-            if ctx.needs_input_grad[2]
-            else None
-        )
-
-        ops.poly_norm_backward(
-            input_grad, weight_grad, bias_grad, output_grad, input, weight, eps
-        )
+        input_grad = torch.empty_like(
+            input) if ctx.needs_input_grad[0] else None
+        weight_grad = torch.empty_like(
+            weight) if ctx.needs_input_grad[1] else None
+        bias_grad = (torch.empty(1, dtype=weight.dtype, device=weight.device)
+                     if ctx.needs_input_grad[2] else None)
+
+        ops.poly_norm_backward(input_grad, weight_grad, bias_grad, output_grad,
+                               input, weight, eps)
 
         return input_grad, weight_grad, bias_grad, None

build/torch27-cxx11-cu128-x86_64-linux/activation/rms_norm.py

@@ -26,9 +26,12 @@ class RMSNormFunction(torch.autograd.Function):
         input, weight = ctx.saved_tensors
         eps = ctx.eps
 
-        input_grad = torch.empty_like(input) if ctx.needs_input_grad[0] else None
-        weight_grad = torch.empty_like(weight) if ctx.needs_input_grad[1] else None
+        input_grad = torch.empty_like(
+            input) if ctx.needs_input_grad[0] else None
+        weight_grad = torch.empty_like(
+            weight) if ctx.needs_input_grad[1] else None
 
-        ops.rms_norm_backward(input_grad, weight_grad, output_grad, input, weight, eps)
+        ops.rms_norm_backward(input_grad, weight_grad, output_grad, input,
+                              weight, eps)
 
         return input_grad, weight_grad, None

build/torch27-cxx11-rocm63-x86_64-linux/activation/{_activation_cf68df1_dirty.abi3.so → _activation_f517c97_dirty.abi3.so}

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ab1037bf6b41bf2be1d00a6a0ed01a97a5e4d64dd0abaf509492ad31eea0a576
-size 2642976
+oid sha256:4be173820e2a4bf4b6b8de6b63faf6544b599d9b0583f650a940adaef4a048b3
+size 2899184

build/torch27-cxx11-rocm63-x86_64-linux/activation/_ops.py

@@ -1,9 +1,9 @@
 import torch
-from . import _activation_cf68df1_dirty
-ops = torch.ops._activation_cf68df1_dirty
+from . import _activation_f517c97_dirty
+ops = torch.ops._activation_f517c97_dirty
 
 def add_op_namespace_prefix(op_name: str):
     """
     Prefix op by namespace.
     """
-    return f"_activation_cf68df1_dirty::{op_name}"
+    return f"_activation_f517c97_dirty::{op_name}"

build/torch27-cxx11-rocm63-x86_64-linux/activation/layers.py

@@ -7,6 +7,7 @@ from .rms_norm import RMSNormFunction
 
 
 class PolyNorm(nn.Module):
+
     def __init__(self, eps=1e-6, dtype: torch.dtype = torch.float32):
         super().__init__()
         self.weight = torch.nn.Parameter(torch.ones(3, dtype=dtype) / 3)
@@ -28,6 +29,7 @@ class PolyNorm(nn.Module):
 
 
 class RMSNorm(nn.Module):
+
     def __init__(self, dim: int, eps=1e-6, dtype: torch.dtype = torch.float32):
         super().__init__()
         self.weight = torch.nn.Parameter(torch.ones(dim, dtype=dtype))

build/torch27-cxx11-rocm63-x86_64-linux/activation/poly_norm.py

@@ -26,16 +26,14 @@ class PolyNormFunction(torch.autograd.Function):
         input, weight = ctx.saved_tensors
         eps = ctx.eps
 
-        input_grad = torch.empty_like(input) if ctx.needs_input_grad[0] else None
-        weight_grad = torch.empty_like(weight) if ctx.needs_input_grad[1] else None
-        bias_grad = (
-            torch.empty(1, dtype=weight.dtype, device=weight.device)
-            if ctx.needs_input_grad[2]
-            else None
-        )
-
-        ops.poly_norm_backward(
-            input_grad, weight_grad, bias_grad, output_grad, input, weight, eps
-        )
+        input_grad = torch.empty_like(
+            input) if ctx.needs_input_grad[0] else None
+        weight_grad = torch.empty_like(
+            weight) if ctx.needs_input_grad[1] else None
+        bias_grad = (torch.empty(1, dtype=weight.dtype, device=weight.device)
+                     if ctx.needs_input_grad[2] else None)
+
+        ops.poly_norm_backward(input_grad, weight_grad, bias_grad, output_grad,
+                               input, weight, eps)
 
         return input_grad, weight_grad, bias_grad, None

build/torch27-cxx11-rocm63-x86_64-linux/activation/rms_norm.py

@@ -26,9 +26,12 @@ class RMSNormFunction(torch.autograd.Function):
         input, weight = ctx.saved_tensors
         eps = ctx.eps
 
-        input_grad = torch.empty_like(input) if ctx.needs_input_grad[0] else None
-        weight_grad = torch.empty_like(weight) if ctx.needs_input_grad[1] else None
+        input_grad = torch.empty_like(
+            input) if ctx.needs_input_grad[0] else None
+        weight_grad = torch.empty_like(
+            weight) if ctx.needs_input_grad[1] else None
 
-        ops.rms_norm_backward(input_grad, weight_grad, output_grad, input, weight, eps)
+        ops.rms_norm_backward(input_grad, weight_grad, output_grad, input,
+                              weight, eps)
 
         return input_grad, weight_grad, None

build/torch28-cxx11-cu126-x86_64-linux/activation/_activation_cf68df1_dirty.abi3.so

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:012788f2064588edf60df24778dff33f8ca95e3b1aaf5243554735cd783dd7ed
-size 3032488

build/torch28-cxx11-cu126-x86_64-linux/activation/_activation_f517c97_dirty.abi3.so

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cb222449350310f90f7271f34fcf9052c9eec28021fee0348130a8f239a97bf4
+size 4571976

build/torch28-cxx11-cu126-x86_64-linux/activation/_ops.py

@@ -1,9 +1,9 @@
 import torch
-from . import _activation_cf68df1_dirty
-ops = torch.ops._activation_cf68df1_dirty
+from . import _activation_f517c97_dirty
+ops = torch.ops._activation_f517c97_dirty
 
 def add_op_namespace_prefix(op_name: str):
     """
     Prefix op by namespace.
     """
-    return f"_activation_cf68df1_dirty::{op_name}"
+    return f"_activation_f517c97_dirty::{op_name}"

build/torch28-cxx11-cu126-x86_64-linux/activation/layers.py

@@ -7,6 +7,7 @@ from .rms_norm import RMSNormFunction
 
 
 class PolyNorm(nn.Module):
+
     def __init__(self, eps=1e-6, dtype: torch.dtype = torch.float32):
         super().__init__()
         self.weight = torch.nn.Parameter(torch.ones(3, dtype=dtype) / 3)
@@ -28,6 +29,7 @@ class PolyNorm(nn.Module):
 
 
 class RMSNorm(nn.Module):
+
     def __init__(self, dim: int, eps=1e-6, dtype: torch.dtype = torch.float32):
         super().__init__()
         self.weight = torch.nn.Parameter(torch.ones(dim, dtype=dtype))

build/torch28-cxx11-cu126-x86_64-linux/activation/poly_norm.py

@@ -26,16 +26,14 @@ class PolyNormFunction(torch.autograd.Function):
         input, weight = ctx.saved_tensors
         eps = ctx.eps
 
-        input_grad = torch.empty_like(input) if ctx.needs_input_grad[0] else None
-        weight_grad = torch.empty_like(weight) if ctx.needs_input_grad[1] else None
-        bias_grad = (
-            torch.empty(1, dtype=weight.dtype, device=weight.device)
-            if ctx.needs_input_grad[2]
-            else None
-        )
-
-        ops.poly_norm_backward(
-            input_grad, weight_grad, bias_grad, output_grad, input, weight, eps
-        )
+        input_grad = torch.empty_like(
+            input) if ctx.needs_input_grad[0] else None
+        weight_grad = torch.empty_like(
+            weight) if ctx.needs_input_grad[1] else None
+        bias_grad = (torch.empty(1, dtype=weight.dtype, device=weight.device)
+                     if ctx.needs_input_grad[2] else None)
+
+        ops.poly_norm_backward(input_grad, weight_grad, bias_grad, output_grad,
+                               input, weight, eps)
 
         return input_grad, weight_grad, bias_grad, None

build/torch28-cxx11-cu126-x86_64-linux/activation/rms_norm.py

@@ -26,9 +26,12 @@ class RMSNormFunction(torch.autograd.Function):
         input, weight = ctx.saved_tensors
         eps = ctx.eps
 
-        input_grad = torch.empty_like(input) if ctx.needs_input_grad[0] else None
-        weight_grad = torch.empty_like(weight) if ctx.needs_input_grad[1] else None
+        input_grad = torch.empty_like(
+            input) if ctx.needs_input_grad[0] else None
+        weight_grad = torch.empty_like(
+            weight) if ctx.needs_input_grad[1] else None
 
-        ops.rms_norm_backward(input_grad, weight_grad, output_grad, input, weight, eps)
+        ops.rms_norm_backward(input_grad, weight_grad, output_grad, input,
+                              weight, eps)
 
         return input_grad, weight_grad, None

build/torch28-cxx11-cu128-x86_64-linux/activation/_activation_cf68df1_dirty.abi3.so

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:b1a65b79b750f550a09e6a1142b5151b03b2a60ec6115a264e6d8de3cac7ee5d
-size 4000920

build/torch28-cxx11-cu128-x86_64-linux/activation/_activation_f517c97_dirty.abi3.so

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:79be6527f579de1133e50a66310d7d0690649dcac63009a54b5e68809408f12a
+size 6634208

build/torch28-cxx11-cu128-x86_64-linux/activation/_ops.py

@@ -1,9 +1,9 @@
 import torch
-from . import _activation_cf68df1_dirty
-ops = torch.ops._activation_cf68df1_dirty
+from . import _activation_f517c97_dirty
+ops = torch.ops._activation_f517c97_dirty
 
 def add_op_namespace_prefix(op_name: str):
     """
     Prefix op by namespace.
     """
-    return f"_activation_cf68df1_dirty::{op_name}"
+    return f"_activation_f517c97_dirty::{op_name}"

build/torch28-cxx11-cu128-x86_64-linux/activation/layers.py

@@ -7,6 +7,7 @@ from .rms_norm import RMSNormFunction
 
 
 class PolyNorm(nn.Module):
+
     def __init__(self, eps=1e-6, dtype: torch.dtype = torch.float32):
         super().__init__()
         self.weight = torch.nn.Parameter(torch.ones(3, dtype=dtype) / 3)
@@ -28,6 +29,7 @@ class PolyNorm(nn.Module):
 
 
 class RMSNorm(nn.Module):
+
     def __init__(self, dim: int, eps=1e-6, dtype: torch.dtype = torch.float32):
         super().__init__()
         self.weight = torch.nn.Parameter(torch.ones(dim, dtype=dtype))

build/torch28-cxx11-cu128-x86_64-linux/activation/poly_norm.py

@@ -26,16 +26,14 @@ class PolyNormFunction(torch.autograd.Function):
         input, weight = ctx.saved_tensors
         eps = ctx.eps
 
-        input_grad = torch.empty_like(input) if ctx.needs_input_grad[0] else None
-        weight_grad = torch.empty_like(weight) if ctx.needs_input_grad[1] else None
-        bias_grad = (
-            torch.empty(1, dtype=weight.dtype, device=weight.device)
-            if ctx.needs_input_grad[2]
-            else None
-        )
-
-        ops.poly_norm_backward(
-            input_grad, weight_grad, bias_grad, output_grad, input, weight, eps
-        )
+        input_grad = torch.empty_like(
+            input) if ctx.needs_input_grad[0] else None
+        weight_grad = torch.empty_like(
+            weight) if ctx.needs_input_grad[1] else None
+        bias_grad = (torch.empty(1, dtype=weight.dtype, device=weight.device)
+                     if ctx.needs_input_grad[2] else None)
+
+        ops.poly_norm_backward(input_grad, weight_grad, bias_grad, output_grad,
+                               input, weight, eps)
 
         return input_grad, weight_grad, bias_grad, None

build/torch28-cxx11-cu128-x86_64-linux/activation/rms_norm.py

@@ -26,9 +26,12 @@ class RMSNormFunction(torch.autograd.Function):
         input, weight = ctx.saved_tensors
         eps = ctx.eps
 
-        input_grad = torch.empty_like(input) if ctx.needs_input_grad[0] else None
-        weight_grad = torch.empty_like(weight) if ctx.needs_input_grad[1] else None
+        input_grad = torch.empty_like(
+            input) if ctx.needs_input_grad[0] else None
+        weight_grad = torch.empty_like(
+            weight) if ctx.needs_input_grad[1] else None
 
-        ops.rms_norm_backward(input_grad, weight_grad, output_grad, input, weight, eps)
+        ops.rms_norm_backward(input_grad, weight_grad, output_grad, input,
+                              weight, eps)
 
         return input_grad, weight_grad, None

build/torch28-cxx11-cu129-x86_64-linux/activation/_activation_cf68df1_dirty.abi3.so

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:fd38039c3401b0f6a136f1761c7f396f5954f05e16d78ed1600d8325c1221781
-size 4059256

build/torch28-cxx11-cu129-x86_64-linux/activation/_activation_f517c97_dirty.abi3.so

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8d95e4491d35cb022a6eaa2febbc555f203893f989a4fb1cc483b2632f141869
+size 6687456

build/torch28-cxx11-cu129-x86_64-linux/activation/_ops.py

@@ -1,9 +1,9 @@
 import torch
-from . import _activation_cf68df1_dirty
-ops = torch.ops._activation_cf68df1_dirty
+from . import _activation_f517c97_dirty
+ops = torch.ops._activation_f517c97_dirty
 
 def add_op_namespace_prefix(op_name: str):
     """
     Prefix op by namespace.
     """
-    return f"_activation_cf68df1_dirty::{op_name}"
+    return f"_activation_f517c97_dirty::{op_name}"

build/torch28-cxx11-cu129-x86_64-linux/activation/layers.py

@@ -7,6 +7,7 @@ from .rms_norm import RMSNormFunction
 
 
 class PolyNorm(nn.Module):
+
     def __init__(self, eps=1e-6, dtype: torch.dtype = torch.float32):
         super().__init__()
         self.weight = torch.nn.Parameter(torch.ones(3, dtype=dtype) / 3)
@@ -28,6 +29,7 @@ class PolyNorm(nn.Module):
 
 
 class RMSNorm(nn.Module):
+
     def __init__(self, dim: int, eps=1e-6, dtype: torch.dtype = torch.float32):
         super().__init__()
         self.weight = torch.nn.Parameter(torch.ones(dim, dtype=dtype))

build/torch28-cxx11-cu129-x86_64-linux/activation/poly_norm.py

@@ -26,16 +26,14 @@ class PolyNormFunction(torch.autograd.Function):
         input, weight = ctx.saved_tensors
         eps = ctx.eps
 
-        input_grad = torch.empty_like(input) if ctx.needs_input_grad[0] else None
-        weight_grad = torch.empty_like(weight) if ctx.needs_input_grad[1] else None
-        bias_grad = (
-            torch.empty(1, dtype=weight.dtype, device=weight.device)
-            if ctx.needs_input_grad[2]
-            else None
-        )
-
-        ops.poly_norm_backward(
-            input_grad, weight_grad, bias_grad, output_grad, input, weight, eps
-        )
+        input_grad = torch.empty_like(
+            input) if ctx.needs_input_grad[0] else None
+        weight_grad = torch.empty_like(
+            weight) if ctx.needs_input_grad[1] else None
+        bias_grad = (torch.empty(1, dtype=weight.dtype, device=weight.device)
+                     if ctx.needs_input_grad[2] else None)
+
+        ops.poly_norm_backward(input_grad, weight_grad, bias_grad, output_grad,
+                               input, weight, eps)
 
         return input_grad, weight_grad, bias_grad, None

build/torch28-cxx11-cu129-x86_64-linux/activation/rms_norm.py

@@ -26,9 +26,12 @@ class RMSNormFunction(torch.autograd.Function):
         input, weight = ctx.saved_tensors
         eps = ctx.eps
 
-        input_grad = torch.empty_like(input) if ctx.needs_input_grad[0] else None
-        weight_grad = torch.empty_like(weight) if ctx.needs_input_grad[1] else None
+        input_grad = torch.empty_like(
+            input) if ctx.needs_input_grad[0] else None
+        weight_grad = torch.empty_like(
+            weight) if ctx.needs_input_grad[1] else None
 
-        ops.rms_norm_backward(input_grad, weight_grad, output_grad, input, weight, eps)
+        ops.rms_norm_backward(input_grad, weight_grad, output_grad, input,
+                              weight, eps)
 
         return input_grad, weight_grad, None

build/torch28-cxx11-rocm63-x86_64-linux/activation/_activation_cf68df1_dirty.abi3.so

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:d8a75fc3e8648bbab973e3021720ed372ec8468f7a28b5b047640fd7198ab369
-size 2647872

build/torch28-cxx11-rocm63-x86_64-linux/activation/_activation_f517c97_dirty.abi3.so

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:58116124bb2b5d11de2753dd0c30a1e4c84759f18599da7016c791bad37528e9
+size 2899984

build/torch28-cxx11-rocm63-x86_64-linux/activation/_ops.py

@@ -1,9 +1,9 @@
 import torch
-from . import _activation_cf68df1_dirty
-ops = torch.ops._activation_cf68df1_dirty
+from . import _activation_f517c97_dirty
+ops = torch.ops._activation_f517c97_dirty
 
 def add_op_namespace_prefix(op_name: str):
     """
     Prefix op by namespace.
     """
-    return f"_activation_cf68df1_dirty::{op_name}"
+    return f"_activation_f517c97_dirty::{op_name}"

build/torch28-cxx11-rocm63-x86_64-linux/activation/layers.py

@@ -7,6 +7,7 @@ from .rms_norm import RMSNormFunction
 
 
 class PolyNorm(nn.Module):
+
     def __init__(self, eps=1e-6, dtype: torch.dtype = torch.float32):
         super().__init__()
         self.weight = torch.nn.Parameter(torch.ones(3, dtype=dtype) / 3)
@@ -28,6 +29,7 @@ class PolyNorm(nn.Module):
 
 
 class RMSNorm(nn.Module):
+
     def __init__(self, dim: int, eps=1e-6, dtype: torch.dtype = torch.float32):
         super().__init__()
         self.weight = torch.nn.Parameter(torch.ones(dim, dtype=dtype))

build/torch28-cxx11-rocm63-x86_64-linux/activation/poly_norm.py

@@ -26,16 +26,14 @@ class PolyNormFunction(torch.autograd.Function):
         input, weight = ctx.saved_tensors
         eps = ctx.eps
 
-        input_grad = torch.empty_like(input) if ctx.needs_input_grad[0] else None
-        weight_grad = torch.empty_like(weight) if ctx.needs_input_grad[1] else None
-        bias_grad = (
-            torch.empty(1, dtype=weight.dtype, device=weight.device)
-            if ctx.needs_input_grad[2]
-            else None
-        )
-
-        ops.poly_norm_backward(
-            input_grad, weight_grad, bias_grad, output_grad, input, weight, eps
-        )
+        input_grad = torch.empty_like(
+            input) if ctx.needs_input_grad[0] else None
+        weight_grad = torch.empty_like(
+            weight) if ctx.needs_input_grad[1] else None
+        bias_grad = (torch.empty(1, dtype=weight.dtype, device=weight.device)
+                     if ctx.needs_input_grad[2] else None)
+
+        ops.poly_norm_backward(input_grad, weight_grad, bias_grad, output_grad,
+                               input, weight, eps)
 
         return input_grad, weight_grad, bias_grad, None

build/torch28-cxx11-rocm63-x86_64-linux/activation/rms_norm.py

@@ -26,9 +26,12 @@ class RMSNormFunction(torch.autograd.Function):
         input, weight = ctx.saved_tensors
         eps = ctx.eps
 
-        input_grad = torch.empty_like(input) if ctx.needs_input_grad[0] else None
-        weight_grad = torch.empty_like(weight) if ctx.needs_input_grad[1] else None
+        input_grad = torch.empty_like(
+            input) if ctx.needs_input_grad[0] else None
+        weight_grad = torch.empty_like(
+            weight) if ctx.needs_input_grad[1] else None
 
-        ops.rms_norm_backward(input_grad, weight_grad, output_grad, input, weight, eps)
+        ops.rms_norm_backward(input_grad, weight_grad, output_grad, input,
+                              weight, eps)
 
         return input_grad, weight_grad, None

build/torch28-cxx11-rocm64-x86_64-linux/activation/_activation_cf68df1_dirty.abi3.so

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:cccb0567a8f86f1f9e23a653a2e1f7177f4528cb1ecf8cbec42e40c60392eb39
-size 2633232

build/torch28-cxx11-rocm64-x86_64-linux/activation/_activation_f517c97_dirty.abi3.so

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:65319d3d93ac3bf0f2939fa4e53ddfc8cd633b9e396cde3a97d63b9041ba03a7
+size 2885344

build/torch28-cxx11-rocm64-x86_64-linux/activation/_ops.py

@@ -1,9 +1,9 @@
 import torch
-from . import _activation_cf68df1_dirty
-ops = torch.ops._activation_cf68df1_dirty
+from . import _activation_f517c97_dirty
+ops = torch.ops._activation_f517c97_dirty
 
 def add_op_namespace_prefix(op_name: str):
     """
     Prefix op by namespace.
     """
-    return f"_activation_cf68df1_dirty::{op_name}"
+    return f"_activation_f517c97_dirty::{op_name}"

build/torch28-cxx11-rocm64-x86_64-linux/activation/layers.py

@@ -7,6 +7,7 @@ from .rms_norm import RMSNormFunction
 
 
 class PolyNorm(nn.Module):
+
     def __init__(self, eps=1e-6, dtype: torch.dtype = torch.float32):
         super().__init__()
         self.weight = torch.nn.Parameter(torch.ones(3, dtype=dtype) / 3)
@@ -28,6 +29,7 @@ class PolyNorm(nn.Module):
 
 
 class RMSNorm(nn.Module):
+
     def __init__(self, dim: int, eps=1e-6, dtype: torch.dtype = torch.float32):
         super().__init__()
         self.weight = torch.nn.Parameter(torch.ones(dim, dtype=dtype))

build/torch28-cxx11-rocm64-x86_64-linux/activation/poly_norm.py

@@ -26,16 +26,14 @@ class PolyNormFunction(torch.autograd.Function):
         input, weight = ctx.saved_tensors
         eps = ctx.eps
 
-        input_grad = torch.empty_like(input) if ctx.needs_input_grad[0] else None
-        weight_grad = torch.empty_like(weight) if ctx.needs_input_grad[1] else None
-        bias_grad = (
-            torch.empty(1, dtype=weight.dtype, device=weight.device)
-            if ctx.needs_input_grad[2]
-            else None
-        )
-
-        ops.poly_norm_backward(
-            input_grad, weight_grad, bias_grad, output_grad, input, weight, eps
-        )
+        input_grad = torch.empty_like(
+            input) if ctx.needs_input_grad[0] else None
+        weight_grad = torch.empty_like(
+            weight) if ctx.needs_input_grad[1] else None
+        bias_grad = (torch.empty(1, dtype=weight.dtype, device=weight.device)
+                     if ctx.needs_input_grad[2] else None)
+
+        ops.poly_norm_backward(input_grad, weight_grad, bias_grad, output_grad,
+                               input, weight, eps)
 
         return input_grad, weight_grad, bias_grad, None

build/torch28-cxx11-rocm64-x86_64-linux/activation/rms_norm.py

@@ -26,9 +26,12 @@ class RMSNormFunction(torch.autograd.Function):
         input, weight = ctx.saved_tensors
         eps = ctx.eps
 
-        input_grad = torch.empty_like(input) if ctx.needs_input_grad[0] else None
-        weight_grad = torch.empty_like(weight) if ctx.needs_input_grad[1] else None
+        input_grad = torch.empty_like(
+            input) if ctx.needs_input_grad[0] else None
+        weight_grad = torch.empty_like(
+            weight) if ctx.needs_input_grad[1] else None
 
-        ops.rms_norm_backward(input_grad, weight_grad, output_grad, input, weight, eps)
+        ops.rms_norm_backward(input_grad, weight_grad, output_grad, input,
+                              weight, eps)
 
         return input_grad, weight_grad, None