Tiny allow DeepGEMM on cu12.9 (#9858)

python/sglang/srt/layers/quantization/deep_gemm_wrapper/configurer.py

@@ -11,9 +11,6 @@ def _compute_enable_deep_gemm():
     sm_version = get_device_sm()
     if sm_version < 90:
         return False
-    # TODO fix deepgemm cu129 fp8 issue
-    if torch.version.cuda == "12.9":
-        return False
 
     try:
         import deep_gemm