EngineX-Hygon/sglang
5
0
Fork 0
Code Issues Pull Requests Actions 7 Projects Releases Wiki Activity

Remove unused imports from rocm mla kernel. (#3963)

Browse Source
This commit is contained in:
Chaitanya Sri Krishna Lolla
2025-02-28 23:31:08 +05:30
committed by GitHub
parent e3e0bc50a9
commit 77a6c9d229
1 changed files with 0 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
python/sglang/srt/layers/attention/triton_ops/rocm_mla_decode_rope.py
View File

@@ -20,19 +20,12 @@ It supports page size = 1.
# https://github.com/ModelTC/lightllm/blob/96353e868a840db4d103138caf15ed9dbea8c186/lightllm/models/deepseek2/triton_kernel/gqa_flash_decoding_stage1.py # https://github.com/ModelTC/lightllm/blob/96353e868a840db4d103138caf15ed9dbea8c186/lightllm/models/deepseek2/triton_kernel/gqa_flash_decoding_stage1.py
# https://github.com/ModelTC/lightllm/blob/96353e868a840db4d103138caf15ed9dbea8c186/lightllm/models/deepseek2/triton_kernel/gqa_flash_decoding_stage2.py # https://github.com/ModelTC/lightllm/blob/96353e868a840db4d103138caf15ed9dbea8c186/lightllm/models/deepseek2/triton_kernel/gqa_flash_decoding_stage2.py
import argparse
import logging
import sys
import pytest
import torch
import triton import triton
import triton.language as tl import triton.language as tl
from sglang.srt.layers.attention.triton_ops.decode_attention import ( from sglang.srt.layers.attention.triton_ops.decode_attention import (
_decode_softmax_reducev_fwd, _decode_softmax_reducev_fwd,
) )
from sglang.srt.layers.rotary_embedding import DeepseekScalingRotaryEmbedding
def is_hip(): def is_hip():