update flash_attn.py

vllm/v1/attention/backends/flash_attn.py

@@ -709,7 +709,7 @@ class FlashAttentionImpl(AttentionImpl):
                     out=output[:num_actual_tokens],
                     cu_seqlens_q=cu_seqlens_q,
                     max_seqlen_q=max_seqlen_q,
-                    seqused_k=seqused_k,
+                    seqused_k=seqused_k, 
                     max_seqlen_k=max_seqlen_k,
                     softmax_scale=self.scale,
                     causal=True,
@@ -717,7 +717,11 @@ class FlashAttentionImpl(AttentionImpl):
                     window_size=self.sliding_window,
                     block_table=block_table,
                     softcap=self.logits_soft_cap,
-                    # scheduler_metadata=scheduler_metadata,
+                    scheduler_metadata=scheduler_metadata,
+                    # fa_version=self.vllm_flash_attn_version,
+                    # q_descale=layer._q_scale.expand(descale_shape),
+                    # k_descale=layer._k_scale.expand(descale_shape),
+                    # v_descale=layer._v_scale.expand(descale_shape),
                 )
             return output