Upload custom kernels
Browse files
build/torch-universal/liger_kernels/_ops.py
CHANGED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
import torch
|
| 2 |
-
ops = torch.ops.
|
| 3 |
|
| 4 |
def add_op_namespace_prefix(op_name: str):
|
| 5 |
"""
|
| 6 |
Prefix op by namespace.
|
| 7 |
"""
|
| 8 |
-
return f"
|
|
|
|
| 1 |
import torch
|
| 2 |
+
ops = torch.ops._liger_kernels_20250507091832
|
| 3 |
|
| 4 |
def add_op_namespace_prefix(op_name: str):
|
| 5 |
"""
|
| 6 |
Prefix op by namespace.
|
| 7 |
"""
|
| 8 |
+
return f"_liger_kernels_20250507091832::{op_name}"
|
build/torch-universal/liger_kernels/layers.py
CHANGED
|
@@ -16,9 +16,6 @@ class LigerRMSNorm(torch.nn.Module):
|
|
| 16 |
|
| 17 |
weight: torch.Tensor
|
| 18 |
variance_epsilon: float
|
| 19 |
-
offset: float = 0
|
| 20 |
-
casting_mode: str = "llama"
|
| 21 |
-
in_place: bool = True
|
| 22 |
|
| 23 |
def forward(self, hidden_states):
|
| 24 |
"""
|
|
@@ -34,9 +31,9 @@ class LigerRMSNorm(torch.nn.Module):
|
|
| 34 |
hidden_states,
|
| 35 |
self.weight,
|
| 36 |
self.variance_epsilon,
|
| 37 |
-
|
| 38 |
-
|
| 39 |
-
|
| 40 |
)
|
| 41 |
|
| 42 |
__all__ = ["LigerRMSNorm"]
|
|
|
|
| 16 |
|
| 17 |
weight: torch.Tensor
|
| 18 |
variance_epsilon: float
|
|
|
|
|
|
|
|
|
|
| 19 |
|
| 20 |
def forward(self, hidden_states):
|
| 21 |
"""
|
|
|
|
| 31 |
hidden_states,
|
| 32 |
self.weight,
|
| 33 |
self.variance_epsilon,
|
| 34 |
+
0,
|
| 35 |
+
"llama",
|
| 36 |
+
True
|
| 37 |
)
|
| 38 |
|
| 39 |
__all__ = ["LigerRMSNorm"]
|
torch-ext/liger_kernels/layers.py
CHANGED
|
@@ -16,9 +16,6 @@ class LigerRMSNorm(torch.nn.Module):
|
|
| 16 |
|
| 17 |
weight: torch.Tensor
|
| 18 |
variance_epsilon: float
|
| 19 |
-
offset: float = 0
|
| 20 |
-
casting_mode: str = "llama"
|
| 21 |
-
in_place: bool = True
|
| 22 |
|
| 23 |
def forward(self, hidden_states):
|
| 24 |
"""
|
|
@@ -34,9 +31,9 @@ class LigerRMSNorm(torch.nn.Module):
|
|
| 34 |
hidden_states,
|
| 35 |
self.weight,
|
| 36 |
self.variance_epsilon,
|
| 37 |
-
|
| 38 |
-
|
| 39 |
-
|
| 40 |
)
|
| 41 |
|
| 42 |
__all__ = ["LigerRMSNorm"]
|
|
|
|
| 16 |
|
| 17 |
weight: torch.Tensor
|
| 18 |
variance_epsilon: float
|
|
|
|
|
|
|
|
|
|
| 19 |
|
| 20 |
def forward(self, hidden_states):
|
| 21 |
"""
|
|
|
|
| 31 |
hidden_states,
|
| 32 |
self.weight,
|
| 33 |
self.variance_epsilon,
|
| 34 |
+
0,
|
| 35 |
+
"llama",
|
| 36 |
+
True
|
| 37 |
)
|
| 38 |
|
| 39 |
__all__ = ["LigerRMSNorm"]
|