Revert "Fix nightly-test CI" (#4065)

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

@@ -427,10 +427,7 @@ class FlashInferAttnBackend(AttentionBackend):
             else:
                 o2, s2 = prefill_wrapper_paged.forward_return_lse(
                     q.contiguous().view(-1, layer.tp_q_head_num, layer.head_dim),
-                    self._to_dtype(
-                        forward_batch.token_to_kv_pool.get_kv_buffer(layer.layer_id),
-                        q.dtype,
-                    ),
+                    forward_batch.token_to_kv_pool.get_kv_buffer(layer.layer_id),
                     causal=False,
                     sm_scale=layer.scaling,
                     logits_soft_cap=layer.logit_cap,
@@ -472,9 +469,7 @@ class FlashInferAttnBackend(AttentionBackend):
 
         o = decode_wrapper.forward(
             q.contiguous().view(-1, layer.tp_q_head_num, layer.head_dim),
-            self._to_dtype(
-                forward_batch.token_to_kv_pool.get_kv_buffer(layer.layer_id), q.dtype
-            ),
+            forward_batch.token_to_kv_pool.get_kv_buffer(layer.layer_id),
             sm_scale=layer.scaling,
             logits_soft_cap=layer.logit_cap,
             k_scale=layer.k_scale,
@@ -483,12 +478,6 @@ class FlashInferAttnBackend(AttentionBackend):
 
         return o.view(-1, layer.tp_q_head_num * layer.head_dim)
 
-    def _to_dtype(self, kv_tuple, dtype):
-        if kv_tuple[0].dtype != dtype:
-            return tuple(t.to(dtype) for t in kv_tuple)
-        else:
-            return kv_tuple
-
     def _get_wrapper_idx(self, layer: RadixAttention):
         if self.num_wrappers == 1:
             return 0