[Qwen4Exp] Support UVA PLE-offload and Engram tensor parallelism (#54371)

Signed-off-by: huanghaoyan.hhy <huanghaoyan.hhy@alibaba-inc.com>
Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
This commit is contained in:
Harry Huang
2026-09-09 07:32:34 -07:00
committed by GitHub
co-authored by mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
parent 62f3bf58a5
commit 3116c5d06b
16 changed files with 1712 additions and 495 deletions
@@ -23,6 +23,9 @@ else:
class QuantizeMethodBase(ABC):
"""Base class for different quantized methods."""
requires_device_loading: bool = True
"""Whether post-load processing requires parameters on the target device."""
uses_meta_device: bool = False
"""
Whether this method creates weights on meta device for online quantization.