[Triton] Fix benchmark_causal.py

benchmarks/benchmark_causal.py

@@ -93,8 +93,8 @@ benchmark_all(flash_attn_unpadded_qkvpacked_func, rearrange(qkv, 'b s ... -> (b
 benchmark_all(attention_pytorch, qkv, dropout_p, causal=causal,
               repeats=repeats, desc='PyTorch Attention')
 
-benchmark_all(flash_attn_qkvpacked_func, qkv, causal=causal, repeats=repeats, desc='FlashAttention Triton')
-pytorch_profiler(flash_attn_qkvpacked_func, qkv, causal=causal, backward=True)
+benchmark_all(flash_attn_qkvpacked_func, qkv, None, causal, repeats=repeats, desc='FlashAttention Triton')
+pytorch_profiler(flash_attn_qkvpacked_func, qkv, None, causal, backward=True)
 
 q, k, v = [torch.randn(batch_size, nheads, seqlen, headdim, device=device, dtype=dtype,
                        requires_grad=True) for _ in range(3)]