Merge branch 'main' into upstream_merge_24_10_21

vllm/attention/backends/rocm_flash_attn.py

@@ -607,18 +607,17 @@ def forward(
             assert attn_metadata.num_encoder_tokens is not None
             num_prefill_tokens = attn_metadata.num_encoder_tokens
 
+        output = torch.empty_like(query)
         # Query for decode. KV is not needed because it is already cached.
         decode_query = query[num_prefill_tokens:]
-
         # QKV for prefill.
         query = query[:num_prefill_tokens]
+
         if key is not None and value is not None:
             key = key[:num_prefill_tokens]
             value = value[:num_prefill_tokens]
 
         if prefill_meta := attn_metadata.prefill_metadata:
-            output = torch.empty_like(query)
-
             # Prompt run.
             # normal attention and DECODER
             if attn_type == AttentionType.DECODER and (
@@ -735,7 +734,6 @@ def forward(
         if decode_meta := attn_metadata.decode_metadata:
             # Decoding run.
             # Whether to use rocm custom paged attention or not
-            output = torch.empty_like(decode_query)
             num_seqs, num_heads, head_size = decode_query.shape
             block_size = value_cache.shape[3]
             gqa_ratio = num_heads // self.num_kv_heads