[Hardware] using current_platform.seed_everything (#9785)

Signed-off-by: wangshuai09 <391746016@qq.com>
This commit is contained in:
wangshuai09
2024-10-29 14:47:44 +00:00
committed by GitHub
parent 09500f7dde
commit 622b7ab955
27 changed files with 104 additions and 105 deletions
+5 -5
View File
@@ -4,7 +4,7 @@ import flashinfer
import pytest
import torch
from vllm.utils import seed_everything
from vllm.platforms import current_platform
NUM_HEADS = [(16, 16), (32, 8), (64, 8), (6, 1)]
HEAD_SIZES = [128, 256]
@@ -84,7 +84,7 @@ def test_flashinfer_decode_with_paged_kv(
soft_cap: Optional[float],
) -> None:
torch.set_default_device("cuda")
seed_everything(0)
current_platform.seed_everything(0)
num_seqs = len(kv_lens)
num_query_heads = num_heads[0]
num_kv_heads = num_heads[1]
@@ -170,7 +170,7 @@ def test_flashinfer_prefill_with_paged_kv(seq_lens: List[Tuple[int, int]],
block_size: int,
soft_cap: Optional[float]) -> None:
torch.set_default_device("cuda")
seed_everything(0)
current_platform.seed_everything(0)
num_seqs = len(seq_lens)
query_lens = [x[0] for x in seq_lens]
kv_lens = [x[1] for x in seq_lens]
@@ -268,7 +268,7 @@ def test_flashinfer_prefill_with_paged_fp8_kv(
head_size: int, dtype: torch.dtype, block_size: int,
soft_cap: Optional[float]) -> None:
torch.set_default_device("cuda")
seed_everything(0)
current_platform.seed_everything(0)
num_seqs = len(seq_lens)
query_lens = [x[0] for x in seq_lens]
kv_lens = [x[1] for x in seq_lens]
@@ -381,7 +381,7 @@ def test_flashinfer_decode_with_paged_fp8_kv(
) -> None:
# test doesn't work for num_heads = (16,16)
torch.set_default_device("cuda")
seed_everything(0)
current_platform.seed_everything(0)
num_seqs = len(kv_lens)
num_query_heads = num_heads[0]
num_kv_heads = num_heads[1]