Fix MLA padding and grouped topk routing in the Transformers modelling backend (#49982)
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com> Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
This commit is contained in:
co-authored by
Claude Opus 5
parent
bf2b45b5d6
commit
99de48e98f
@@ -16,7 +16,9 @@
|
||||
# limitations under the License.
|
||||
"""Transformers modeling backend utilities."""
|
||||
|
||||
from collections.abc import Iterator
|
||||
from contextlib import contextmanager
|
||||
from itertools import chain
|
||||
from pathlib import Path
|
||||
from typing import TYPE_CHECKING, Literal
|
||||
|
||||
@@ -209,6 +211,11 @@ def recursive_replace_linear(
|
||||
_recursive_replace(model, prefix=prefix)
|
||||
|
||||
|
||||
def named_state(module: nn.Module) -> Iterator[tuple[str, torch.Tensor]]:
|
||||
"""`module`'s own state (i.e. named parameters and buffers)."""
|
||||
return chain(module.named_parameters(), module.named_buffers())
|
||||
|
||||
|
||||
def log_replacement(name: str, old_module: nn.Module, new_module: nn.Module):
|
||||
logger.debug("%s: %s -> %s", name, old_module, new_module)
|
||||
|
||||
|
||||
Reference in New Issue
Block a user