Files
xc-llm-ascend/vllm_ascend
Debonet 6ce1dc162a [v0.18.0] fix(attention): reuse weight address in graph + RL scenario (#7715)
### What this PR does / why we need it?

In graph + RL scenario, we only capture the graph once, and the weight
address is expected to be the same across iterations. However, when
calling .contiguous() on weight tensors, a new memory address may be
allocated, causing the graph to capture incorrect weight addresses.
This PR modifies the weight update logic in AscendMLAImpl and
AscendSFAImpl to use copy_() instead of reassignment, ensuring the
weight addresses remain consistent across iterations.

detailed in #7473

### Does this PR introduce _any_ user-facing change?

### How was this patch tested?

Signed-off-by: Debonex <719893090@qq.com>
2026-03-27 14:11:20 +08:00
..
2026-03-21 16:05:38 +08:00
2026-03-19 14:27:27 +08:00