stick to regular 2:4 sparsity, avoid marlin kernel

Signed-off-by: Brian Dellabetta <[email protected]>
vllm-project · Feb 10, 2025 · 7b39550 · 7b39550
1 parent 3462e17
commit 7b39550
Show file tree

Hide file tree

Showing 3 changed files with 8 additions and 18 deletions.
diff --git a/tests/e2e/vLLM/configs/sparse_24_qwen.yaml b/tests/e2e/vLLM/configs/sparse_24_qwen.yaml
@@ -0,0 +1,8 @@
+cadence: "nightly"
+test_type: "regression"
+model: Qwen/Qwen2.5-0.5B
+recipe: tests/e2e/vLLM/recipes/Sparse_2of4/recipe_sparse_2of4.yaml
+scheme: sparse2of4_only
+dataset_id: garage-bAInd/Open-Platypus
+dataset_split: train
+save_compressed: False
diff --git a/tests/e2e/vLLM/configs/w4a16_2of4_channel_quant_qwen.yaml b/tests/e2e/vLLM/configs/w4a16_2of4_channel_quant_qwen.yaml
diff --git a/tests/e2e/vLLM/configs/w4a16_2of4_grouped_quant_qwen.yaml b/tests/e2e/vLLM/configs/w4a16_2of4_grouped_quant_qwen.yaml