Making large AI models cheaper, faster and more accessible
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
 
 
 
 
 

76 lines
2.3 KiB

import torch
import triton
from vllm._C import ops
from colossalai.kernel.kernel_loader import InferenceOpsLoader
from colossalai.kernel.triton import rotary_embedding
inference_ops = InferenceOpsLoader().load()
BATCH = 16
configs = [
triton.testing.Benchmark(
x_names=["num_tokens"],
x_vals=[2**i for i in range(4, 12)],
line_arg="provider",
line_vals=["triton_func", "colossal_cuda_func", "vllm_cuda_func"],
line_names=["triton_func", "colossal_cuda_func", "vllm_cuda_func"],
styles=[("red", "-"), ("blue", "-"), ("yellow", "-")],
ylabel="ms",
plot_name=f"rotary_emb-batch-{BATCH}",
args={"num_kv_heads": 16},
)
]
def torch_rotary_emb(x, cos, sin):
seq_len, h, dim = x.shape
x0 = x[:, :, 0 : dim // 2]
x1 = x[:, :, dim // 2 : dim]
cos = cos.view((seq_len, 1, dim // 2))
sin = sin.view((seq_len, 1, dim // 2))
o0 = x0 * cos - x1 * sin
o1 = x0 * sin + x1 * cos
return torch.cat((o0, o1), dim=-1)
@triton.testing.perf_report(configs)
def benchmark_rotary_emb(
provider: str,
num_tokens: int,
num_kv_heads: int,
):
warmup = 10
rep = 100
head_dim = 128
dtype = torch.float16
q_shape = (num_tokens, num_kv_heads, head_dim)
q = -2.3 + 0.5 * torch.randn(q_shape, dtype=dtype, device="cuda")
k_shape = (num_tokens, num_kv_heads, head_dim)
k = -2.3 + 0.5 * torch.randn(k_shape, dtype=dtype, device="cuda")
cos_shape = (4096, head_dim // 2)
cos = -1.2 + 0.5 * torch.randn(cos_shape, dtype=dtype, device="cuda")
sin = -2.0 + 0.5 * torch.randn(cos_shape, dtype=dtype, device="cuda")
cos_sin = torch.stack((cos, sin), dim=1).contiguous()
positions = torch.arange(num_tokens).cuda()
if provider == "triton_func":
fn = lambda: rotary_embedding(q, k, cos, sin)
elif provider == "colossal_cuda_func":
fn = lambda: inference_ops.rotary_embedding(q, k, cos, sin)
elif provider == "vllm_cuda_func":
q = q.view(num_tokens, -1)
k = k.view(num_tokens, -1)
fn = lambda: ops.rotary_embedding(positions, q, k, head_dim, cos_sin, True)
else:
raise ValueError("Undefined provider")
ms = triton.testing.do_bench(fn, warmup=warmup, rep=rep)
return ms
if __name__ == "__main__":
benchmark_rotary_emb.run(save_path=".", print_data=True)