[Attention][FlashInfer] Enable FP8 FlashInfer (TRTLLM) MLA decode (#24705)
Signed-off-by: Matthew Bonanni <mbonanni001@gmail.com>
This commit is contained in:
@@ -178,6 +178,7 @@ class MockAttentionLayer:
|
||||
self._k_scale = torch.tensor(1.0, device=device)
|
||||
self._v_scale = torch.tensor(1.0, device=device)
|
||||
# Add float versions for flashinfer
|
||||
self._q_scale_float = 1.0
|
||||
self._k_scale_float = 1.0
|
||||
self._v_scale_float = 1.0
|
||||
|
||||
|
||||
@@ -33,10 +33,12 @@ def test_ragged_paged_attention():
|
||||
)
|
||||
|
||||
class FakeAttentionLayer:
|
||||
_q_scale_float: float
|
||||
_k_scale_float: float
|
||||
_v_scale_float: float
|
||||
|
||||
layer = FakeAttentionLayer()
|
||||
layer._q_scale_float = 1.0
|
||||
layer._k_scale_float = 1.0
|
||||
layer._v_scale_float = 1.0
|
||||
|
||||
|
||||
Reference in New Issue
Block a user