Revert "[bugfix] Fix PPMissingLayer AttributeError when Using PP" (#21189)
This commit is contained in:
@@ -59,28 +59,12 @@ class PPMissingLayer(torch.nn.Identity):
|
|||||||
# https://github.com/vllm-project/vllm/blob/18ed3132d2bfe1df9a74729457b69243955221e8/vllm/model_executor/models/utils.py#L468C1-L486C1
|
# https://github.com/vllm-project/vllm/blob/18ed3132d2bfe1df9a74729457b69243955221e8/vllm/model_executor/models/utils.py#L468C1-L486C1
|
||||||
"""
|
"""
|
||||||
A placeholder layer for missing layers in a pipeline parallel model.
|
A placeholder layer for missing layers in a pipeline parallel model.
|
||||||
|
|
||||||
Attribute access on this placeholder returns 'self' so that common
|
|
||||||
patterns like ``isinstance(layer.mlp, SomeMoE)`` safely evaluate to
|
|
||||||
``False`` instead of raising ``AttributeError``.
|
|
||||||
"""
|
"""
|
||||||
|
|
||||||
def __init__(self, *args, **kwargs):
|
def __init__(self, *args, **kwargs):
|
||||||
super().__init__()
|
super().__init__()
|
||||||
self.return_tuple = kwargs.get("return_tuple", False)
|
self.return_tuple = kwargs.get("return_tuple", False)
|
||||||
|
|
||||||
def __getattr__(self, name: str):
|
|
||||||
try:
|
|
||||||
return super().__getattr__(name)
|
|
||||||
except (AttributeError, KeyError):
|
|
||||||
return self
|
|
||||||
|
|
||||||
def __setattr__(self, name: str, value):
|
|
||||||
if name == "return_tuple" or name.startswith("_"):
|
|
||||||
super().__setattr__(name, value)
|
|
||||||
else:
|
|
||||||
pass
|
|
||||||
|
|
||||||
def forward(self, *args, **kwargs):
|
def forward(self, *args, **kwargs):
|
||||||
"""
|
"""
|
||||||
Return the first arg from args or the first value from kwargs.
|
Return the first arg from args or the first value from kwargs.
|
||||||
|
|||||||
Reference in New Issue
Block a user