@@ -86,8 +86,8 @@ def wrapper(*args, **kwargs):
8686# @pytest.mark.parametrize('d', [32, 40, 64, 80, 96, 128])
8787# @pytest.mark.parametrize("d", [64, 96, 128, 192])
8888# @pytest.mark.parametrize("d", [128, 192])
89- @pytest .mark .parametrize ("d" , [64 , 96 , 128 , 256 ])
90- # @pytest.mark.parametrize("d", [192 ])
89+ @pytest .mark .parametrize ("d" , [64 , 96 , 128 , 192 , 256 ])
90+ # @pytest.mark.parametrize("d", [128 ])
9191@pytest .mark .parametrize (
9292 "seqlen_q,seqlen_k" ,
9393 [
@@ -493,8 +493,8 @@ def test_flash_attn_small_head_dim(seqlen_q, seqlen_k, d, causal, dtype):
493493# @pytest.mark.parametrize("deterministic", [False])
494494@pytest .mark .parametrize ("softcap" , [0.0 , 15.0 ])
495495# @pytest.mark.parametrize("softcap", [0.0])
496- # @pytest.mark.parametrize("local_enum", [0, 1, 2, 3])
497- @pytest .mark .parametrize ("local_enum" , [0 ])
496+ @pytest .mark .parametrize ("local_enum" , [0 , 1 , 2 , 3 ])
497+ # @pytest.mark.parametrize("local_enum", [0])
498498@pytest .mark .parametrize ("causal" , [False , True ])
499499# @pytest.mark.parametrize("causal", [False])
500500# @pytest.mark.parametrize("add_unused_qkv", [False, True])
@@ -505,8 +505,8 @@ def test_flash_attn_small_head_dim(seqlen_q, seqlen_k, d, causal, dtype):
505505# @pytest.mark.parametrize('d', [56, 80])
506506# @pytest.mark.parametrize('d', [32, 40, 64, 80, 96, 128])
507507# @pytest.mark.parametrize("d", [64, 96, 128])
508- @pytest .mark .parametrize ("d" , [128 , 192 ])
509- # @pytest.mark.parametrize("d", [64, 128, 192, 256])
508+ # @pytest.mark.parametrize("d", [128, 192])
509+ @pytest .mark .parametrize ("d" , [64 , 128 , 192 , 256 ])
510510@pytest .mark .parametrize (
511511 "seqlen_q,seqlen_k" ,
512512 [
@@ -519,13 +519,13 @@ def test_flash_attn_small_head_dim(seqlen_q, seqlen_k, d, causal, dtype):
519519 (128 , 128 ),
520520 (256 , 256 ),
521521 (113 , 203 ),
522- # (128, 217),
523- # (113, 211),
524- # (108, 256),
525- # (256, 512),
526- # (307, 256),
527- # (640, 128),
528- # (512, 256),
522+ (128 , 217 ),
523+ (113 , 211 ),
524+ (108 , 256 ),
525+ (256 , 512 ),
526+ (307 , 256 ),
527+ (640 , 128 ),
528+ (512 , 256 ),
529529 (1024 , 1024 ),
530530 (1023 , 1024 ),
531531 (1024 , 1023 ),
@@ -537,8 +537,8 @@ def test_flash_attn_small_head_dim(seqlen_q, seqlen_k, d, causal, dtype):
537537 (4224 , 4224 ),
538538 ],
539539)
540- # @pytest.mark.parametrize("varlen_mode", ["random", "third", "full"])
541- @pytest .mark .parametrize ("varlen_mode" , ["full" ])
540+ @pytest .mark .parametrize ("varlen_mode" , ["random" , "third" , "full" ])
541+ # @pytest.mark.parametrize("varlen_mode", ["full"])
542542@pytest .mark .parametrize (
543543 "zero_lengths_q, zero_lengths_k" ,
544544 [
0 commit comments