Revert "[Misc] Add FA2 support to ViT MHA layer (#12355)"

This reverts commit f1fc051.
vllm-project · Jan 26, 2025 · 1be4877 · 1be4877
1 parent 324960a
commit 1be4877
Show file tree

Hide file tree

Showing 2 changed files with 5 additions and 146 deletions.
diff --git a/tests/kernels/test_mha_attn.py b/tests/kernels/test_mha_attn.py
diff --git a/vllm/attention/layer.py b/vllm/attention/layer.py
@@ -210,22 +210,18 @@ def __init__(
         self.scale = scale
         self.num_kv_heads = num_heads if num_kv_heads is None else num_kv_heads
 
-        assert self.num_heads % self.num_kv_heads == 0
-        self.num_queries_per_kv = self.num_heads // self.num_kv_heads
-
         dtype = torch.get_default_dtype()
         attn_backend = get_attn_backend(head_size,
                                         dtype,
                                         kv_cache_dtype=None,
                                         block_size=16,
                                         is_attention_free=False)
         backend = backend_name_to_enum(attn_backend.get_name())
+        if backend in {_Backend.FLASH_ATTN, _Backend.FLASH_ATTN_VLLM_V1}:
+            backend = _Backend.XFORMERS
 
         self.attn_backend = backend if backend in {
-            _Backend.TORCH_SDPA,
-            _Backend.XFORMERS,
-            _Backend.FLASH_ATTN,
-            _Backend.FLASH_ATTN_VLLM_V1,
+            _Backend.TORCH_SDPA, _Backend.XFORMERS
         } else _Backend.TORCH_SDPA
 
     def forward(
@@ -235,26 +231,15 @@ def forward(
         value: torch.Tensor,
     ) -> torch.Tensor:
         """Input shape: batch_size x seq_len x hidden_size"""
+        # TODO(Isotr0py): Use existing backend implementations and support FA2
         bsz, q_len, _ = query.size()
         kv_len = key.size(1)
 
         query = query.view(bsz, q_len, self.num_heads, self.head_size)
         key = key.view(bsz, kv_len, self.num_kv_heads, self.head_size)
         value = value.view(bsz, kv_len, self.num_kv_heads, self.head_size)
 
-        if (num_repeat := self.num_queries_per_kv) > 1:
-            # Handle MQA and GQA
-            key = torch.repeat_interleave(key, num_repeat, dim=2)
-            value = torch.repeat_interleave(value, num_repeat, dim=2)
-
-        if self.attn_backend in {
-                _Backend.FLASH_ATTN,
-                _Backend.FLASH_ATTN_VLLM_V1,
-        }:
-            from vllm.vllm_flash_attn import flash_attn_func
-
-            out = flash_attn_func(query, key, value, softmax_scale=self.scale)
-        elif self.attn_backend == _Backend.XFORMERS:
+        if self.attn_backend == _Backend.XFORMERS:
             from xformers import ops as xops
 
             out = xops.memory_efficient_attention_forward(query,