[Chore] Separate out optional dependency checks from vllm.utils (#27207)
Signed-off-by: dongbo910220 <1275604947@qq.com> Signed-off-by: dongbo910220 <32610838+dongbo910220@users.noreply.github.com> Co-authored-by: Wentao Ye <44945378+yewentao256@users.noreply.github.com> Co-authored-by: Cyrus Leung <tlleungac@connect.ust.hk>
This commit is contained in:
co-authored by
Wentao Ye
Cyrus Leung
parent
49c00fe304
commit
3ae082c373
@@ -6,7 +6,7 @@ import pytest
|
||||
import torch
|
||||
|
||||
from vllm.platforms import current_platform
|
||||
from vllm.utils import cdiv, has_deep_gemm
|
||||
from vllm.utils import cdiv
|
||||
from vllm.utils.deep_gemm import (
|
||||
_ceil_to_ue8m0,
|
||||
calc_diff,
|
||||
@@ -15,6 +15,7 @@ from vllm.utils.deep_gemm import (
|
||||
get_num_sms,
|
||||
get_paged_mqa_logits_metadata,
|
||||
)
|
||||
from vllm.utils.import_utils import has_deep_gemm
|
||||
|
||||
|
||||
def kv_cache_cast_to_fp8(x: torch.Tensor) -> torch.Tensor:
|
||||
|
||||
@@ -23,7 +23,7 @@ from vllm.model_executor.layers.fused_moe.config import (
|
||||
FusedMoEQuantConfig,
|
||||
)
|
||||
from vllm.model_executor.layers.fused_moe.fused_moe import fused_topk
|
||||
from vllm.utils import has_deep_ep, has_deep_gemm, has_pplx
|
||||
from vllm.utils.import_utils import has_deep_ep, has_deep_gemm, has_pplx
|
||||
|
||||
from .mk_objects import (
|
||||
TestMoEQuantConfig,
|
||||
|
||||
@@ -35,9 +35,9 @@ from vllm.model_executor.layers.quantization.utils.w8a8_utils import (
|
||||
cutlass_fp8_supported,
|
||||
)
|
||||
from vllm.platforms import current_platform
|
||||
from vllm.utils import has_deep_ep, has_deep_gemm, has_pplx
|
||||
from vllm.utils.deep_gemm import is_deep_gemm_supported
|
||||
from vllm.utils.flashinfer import has_flashinfer_cutlass_fused_moe
|
||||
from vllm.utils.import_utils import has_deep_ep, has_deep_gemm, has_pplx
|
||||
|
||||
|
||||
@dataclass
|
||||
|
||||
@@ -15,7 +15,7 @@ from torch.distributed import ProcessGroup
|
||||
from torch.multiprocessing import spawn # pyright: ignore[reportPrivateImportUsage]
|
||||
from typing_extensions import ParamSpec
|
||||
|
||||
from vllm.utils import has_deep_ep
|
||||
from vllm.utils.import_utils import has_deep_ep
|
||||
from vllm.utils.network_utils import get_open_port
|
||||
|
||||
if has_deep_ep():
|
||||
|
||||
@@ -21,11 +21,11 @@ from vllm.model_executor.layers.fused_moe.fused_moe import (
|
||||
modular_triton_fused_moe,
|
||||
)
|
||||
from vllm.platforms import current_platform
|
||||
from vllm.utils import has_deep_gemm
|
||||
from vllm.utils.deep_gemm import (
|
||||
get_mk_alignment_for_contiguous_layout,
|
||||
is_deep_gemm_e8m0_used,
|
||||
)
|
||||
from vllm.utils.import_utils import has_deep_gemm
|
||||
|
||||
dg_available = has_deep_gemm()
|
||||
|
||||
|
||||
@@ -21,8 +21,8 @@ from vllm.model_executor.layers.fused_moe.config import (
|
||||
from vllm.model_executor.layers.fused_moe.fused_moe import fused_experts
|
||||
from vllm.model_executor.layers.fused_moe.modular_kernel import FusedMoEModularKernel
|
||||
from vllm.platforms import current_platform
|
||||
from vllm.utils import has_deep_ep, has_deep_gemm
|
||||
from vllm.utils.deep_gemm import is_deep_gemm_e8m0_used, is_deep_gemm_supported
|
||||
from vllm.utils.import_utils import has_deep_ep, has_deep_gemm
|
||||
|
||||
from ...utils import multi_gpu_test
|
||||
from .parallel_utils import ProcessGroupInfo, parallel_launch
|
||||
|
||||
@@ -21,7 +21,7 @@ from vllm.model_executor.layers.quantization.utils.fp8_utils import (
|
||||
per_token_group_quant_fp8,
|
||||
)
|
||||
from vllm.platforms import current_platform
|
||||
from vllm.utils import has_deep_ep
|
||||
from vllm.utils.import_utils import has_deep_ep
|
||||
|
||||
from ...utils import multi_gpu_test
|
||||
from .parallel_utils import ProcessGroupInfo, parallel_launch
|
||||
|
||||
@@ -6,7 +6,7 @@ import pytest
|
||||
import torch
|
||||
import torch.nn.functional as F
|
||||
|
||||
from vllm.utils import has_triton_kernels
|
||||
from vllm.utils.import_utils import has_triton_kernels
|
||||
|
||||
if not has_triton_kernels():
|
||||
pytest.skip(
|
||||
|
||||
@@ -13,8 +13,8 @@ import torch
|
||||
import vllm.model_executor.layers.fused_moe.modular_kernel as mk
|
||||
from vllm.config import VllmConfig, set_current_vllm_config
|
||||
from vllm.platforms import current_platform
|
||||
from vllm.utils import has_deep_ep, has_deep_gemm, has_pplx
|
||||
from vllm.utils.flashinfer import has_flashinfer_cutlass_fused_moe
|
||||
from vllm.utils.import_utils import has_deep_ep, has_deep_gemm, has_pplx
|
||||
from vllm.utils.torch_utils import cuda_device_count_stateless
|
||||
|
||||
from .modular_kernel_tools.common import (
|
||||
|
||||
@@ -18,12 +18,12 @@ from vllm.model_executor.layers.quantization.utils.fp8_utils import (
|
||||
w8a8_triton_block_scaled_mm,
|
||||
)
|
||||
from vllm.platforms import current_platform
|
||||
from vllm.utils import has_deep_gemm
|
||||
from vllm.utils.deep_gemm import (
|
||||
fp8_gemm_nt,
|
||||
get_col_major_tma_aligned_tensor,
|
||||
per_block_cast_to_fp8,
|
||||
)
|
||||
from vllm.utils.import_utils import has_deep_gemm
|
||||
|
||||
if current_platform.get_device_capability() < (9, 0):
|
||||
pytest.skip("FP8 Triton requires CUDA 9.0 or higher", allow_module_level=True)
|
||||
|
||||
Reference in New Issue
Block a user