biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

Lora MoE Align Improvements (#29257)

Browse Source 
Signed-off-by: gnovack <gnovack@amazon.com>
This commit is contained in:
gnovack
2025-12-08 18:35:16 -08:00
committed by GitHub
parent db14f61f2d
commit ea657f2078
7 changed files with 360 additions and 249 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
tests/lora/test_moe_lora_align_sum.py
View File

@@ -32,7 +32,7 @@ def sample_data(num_experts, max_loras, num_tokens, topk_num):
@pytest.mark.parametrize("num_tokens", [100, 200, 1024, 4096]) # 81920
@pytest.mark.parametrize("topk_num", [6])
@pytest.mark.parametrize("num_experts", [64, 128])
@pytest.mark.parametrize("num_experts", [64, 128, 256, 512])
@pytest.mark.parametrize("max_loras", [2, 32])
@pytest.mark.parametrize("block_size", [16])
def test_moe_lora_align_block_size(