Add FP8 Blockwise GEMM Backend Flag --fp8-gemm-backend (#14379)

2025-12-09 12:05:56 -08:00
parent 73df7a4e8d
commit 55504df2f7
10 changed files with 306 additions and 33 deletions
@@ -170,6 +170,7 @@ suites = {
    "per-commit-4-gpu-b200": [
        TestFile("test_deepseek_v3_fp4_4gpu.py", 1800),
        TestFile("test_flash_attention_4.py", 300),
+        TestFile("test_fp8_blockwise_gemm.py", 280),
        TestFile("test_gpt_oss_4gpu.py", 600),
        TestFile("test_llama31_fp4.py", 300),
        TestFile("test_eagle_infer_beta_dp_attention.py", 300),