support fp8_e4m3.

python/sglang/srt/layers/attention/dcu_mla_backend.py

@@ -465,9 +465,9 @@ class DCUMLABackend(AttentionBackend):
             getattr(torch, "float8_e5m2fnuz", None),
         ):
             if k_cache_reshaped.dtype == torch.float8_e4m3fnuz:
-                k_cache_reshaped = k_cache_reshaped.view(torch.float8_e4m3fn)
+                kv_cache_dtype="fp8_e4m3"
             elif k_cache_reshaped.dtype == torch.float8_e5m2fnuz:
-                k_cache_reshaped = k_cache_reshaped.view(torch.float8_e5m2)
+                kv_cache_dtype="fp8_e5m2"
             k_scale = layer.k_scale if layer.k_scale is not None else torch.tensor([1.0], dtype=torch.float32, device=reshape_q.device)
             o = self._call_fp8_decode(
                 reshape_q, 
@@ -476,7 +476,7 @@ class DCUMLABackend(AttentionBackend):
                 (forward_batch.seq_lens + self.num_draft_tokens).to(torch.int32),
                 layer.scaling,
                 k_scale.to(torch.float32),
-                kv_cache_dtype=self.data_type,
+                kv_cache_dtype=kv_cache_dtype,
             )
         else:
             o = self._call_decode(