Ruri
aff5189c87
[main] Fuse GroupedMatmul, Swiglu and DynamicQuant in W8A8_DYNAMIC quantized MoE layers (#2275)
### What this PR does / why we need it?
Fuse `GroupedMatmul`, `Swiglu` and `DynamicQuant` into one fusion
operation `GroupedMatmulSwigluQuant`.
1. extract common functions in `w4a8_dynamic.py` and `w8a8_dynamic.py`
2. if in supported occasion, use fusion operation
`npu_grouped_matmul_swiglu_quant`
### Does this PR introduce _any_ user-facing change?
### How was this patch tested?
Tested on W8A8 quantized Qwen3-235B-A22B model with `bs=16`
1. `tp=8`, `dp=1`, `moe_tp=8`, `moe_ep=1`, TPOP increased 21.54%, Output
Token Throughput increased 27.35%
<img width="3443" height="211" alt="image"
src="https://github.com/user-attachments/assets/a1a9c14d-2310-41be-9a03-36125dabae6e"
/>
3. `tp=8`, `dp=1`, `moe_tp=1`, `moe_ep=8`, TPOP increased 17.38%, Output
Token Throughput increased 6.86%
<img width="3443" height="211" alt="image"
src="https://github.com/user-attachments/assets/1ce92e92-720d-40c0-8b4d-c493e5cb10a6"
/>
- vLLM version: v0.10.1.1
- vLLM main:
6997a25ac6
---------
Signed-off-by: Ruri <33858552+zhoux77899@users.noreply.github.com>
Signed-off-by: zhoux77899 <zhouxiang100@huawei.com>
2025-09-04 11:37:32 +08:00
..
2025-09-04 10:22:46 +08:00
2025-08-30 12:04:01 +08:00
2025-07-31 19:17:27 +08:00
2025-08-29 11:41:21 +08:00
2025-06-16 18:32:28 +08:00
2025-09-04 08:22:10 +08:00
2025-09-03 14:38:55 +08:00
2025-07-31 19:17:27 +08:00
2025-09-04 11:37:32 +08:00
2025-08-14 09:33:39 +08:00
2025-09-04 11:37:32 +08:00
2025-08-11 17:37:49 +08:00
2025-09-04 11:35:14 +08:00
2025-09-03 10:58:08 +08:00
2025-07-21 19:43:30 +08:00
2025-07-28 15:13:37 +08:00
2025-07-18 23:07:14 +08:00
2025-09-03 17:56:12 +08:00
2025-08-14 09:33:39 +08:00
2025-09-02 18:34:04 +08:00
2025-08-29 11:41:21 +08:00