94 lines
2.8 KiB
Python
94 lines
2.8 KiB
Python
import pytest
|
|
import torch
|
|
|
|
from vllm.model_executor.layers.activation import FastGELU, NewGELU, SiluAndMul
|
|
from allclose_default import get_default_atol, get_default_rtol
|
|
|
|
DTYPES = [torch.half, torch.bfloat16, torch.float]
|
|
NUM_TOKENS = [7, 83, 2048] # Arbitrary values for testing
|
|
D = [512, 4096, 5120, 13824] # Arbitrary values for testing
|
|
SEEDS = [0]
|
|
CUDA_DEVICES = [
|
|
f"cuda:{i}" for i in range(1 if torch.cuda.device_count() == 1 else 2)
|
|
]
|
|
|
|
|
|
@pytest.mark.parametrize("num_tokens", NUM_TOKENS)
|
|
@pytest.mark.parametrize("d", D)
|
|
@pytest.mark.parametrize("dtype", DTYPES)
|
|
@pytest.mark.parametrize("seed", SEEDS)
|
|
@pytest.mark.parametrize("device", CUDA_DEVICES)
|
|
@torch.inference_mode()
|
|
def test_silu_and_mul(
|
|
num_tokens: int,
|
|
d: int,
|
|
dtype: torch.dtype,
|
|
seed: int,
|
|
device: str,
|
|
) -> None:
|
|
torch.random.manual_seed(seed)
|
|
if torch.cuda.is_available():
|
|
torch.cuda.manual_seed(seed)
|
|
torch.set_default_device(device)
|
|
x = torch.randn(num_tokens, 2 * d, dtype=dtype)
|
|
layer = SiluAndMul()
|
|
out = layer(x)
|
|
ref_out = layer._forward(x)
|
|
assert torch.allclose(out,
|
|
ref_out,
|
|
atol=get_default_atol(out),
|
|
rtol=get_default_rtol(out))
|
|
|
|
|
|
@pytest.mark.parametrize("num_tokens", NUM_TOKENS)
|
|
@pytest.mark.parametrize("d", D)
|
|
@pytest.mark.parametrize("dtype", DTYPES)
|
|
@pytest.mark.parametrize("seed", SEEDS)
|
|
@pytest.mark.parametrize("device", CUDA_DEVICES)
|
|
@torch.inference_mode()
|
|
def test_gelu_new(
|
|
num_tokens: int,
|
|
d: int,
|
|
dtype: torch.dtype,
|
|
seed: int,
|
|
device: str,
|
|
) -> None:
|
|
torch.random.manual_seed(seed)
|
|
if torch.cuda.is_available():
|
|
torch.cuda.manual_seed(seed)
|
|
torch.set_default_device(device)
|
|
x = torch.randn(num_tokens, d, dtype=dtype)
|
|
layer = NewGELU()
|
|
out = layer(x)
|
|
ref_out = layer._forward(x)
|
|
assert torch.allclose(out,
|
|
ref_out,
|
|
atol=get_default_atol(out),
|
|
rtol=get_default_rtol(out))
|
|
|
|
|
|
@pytest.mark.parametrize("num_tokens", NUM_TOKENS)
|
|
@pytest.mark.parametrize("d", D)
|
|
@pytest.mark.parametrize("dtype", DTYPES)
|
|
@pytest.mark.parametrize("seed", SEEDS)
|
|
@pytest.mark.parametrize("device", CUDA_DEVICES)
|
|
def test_gelu_fast(
|
|
num_tokens: int,
|
|
d: int,
|
|
dtype: torch.dtype,
|
|
seed: int,
|
|
device: str,
|
|
) -> None:
|
|
torch.random.manual_seed(seed)
|
|
if torch.cuda.is_available():
|
|
torch.cuda.manual_seed(seed)
|
|
torch.set_default_device(device)
|
|
x = torch.randn(num_tokens, d, dtype=dtype)
|
|
layer = FastGELU()
|
|
out = layer(x)
|
|
ref_out = layer._forward(x)
|
|
assert torch.allclose(out,
|
|
ref_out,
|
|
atol=get_default_atol(out),
|
|
rtol=get_default_rtol(out))
|