feat: add sonicmoe fused lora support (#3519)

* feat: add sonicmoe fused lora support

* fix: forgot to add file

* feat: add test

* feat: add lora support for other routes

* fix: add int8 lora support

* fix: add qwen35_moe interleave support

* fix: qwen3_5_moe loss

* chore: lint

* address some pr comments

* fix test imports

* add support matrix for moe kernels [skip ci]

---------

Co-authored-by: Wing Lian <wing@axolotl.ai>
This commit is contained in:
NanoCode012
2026-04-02 19:53:48 +07:00
committed by GitHub
parent 16e32232fb
commit 842fa039dd
16 changed files with 1249 additions and 126 deletions

View File

@@ -7,7 +7,7 @@ code path where routing happens in float32.
import torch
from axolotl.integrations.kernels.sonicmoe.routing import (
from axolotl.integrations.kernels.libs.sonicmoe.routing import (
sigmoid_topk_routing,
softmax_topk_routing,
)