[Chore]:Extract math and argparse utilities to separate modules (#27188)
Signed-off-by: Yeshwanth Surya <yeshsurya@gmail.com> Signed-off-by: Yeshwanth N <yeshsurya@gmail.com> Signed-off-by: yeshsurya <yeshsurya@gmail.com>
This commit is contained in:
@@ -5,7 +5,7 @@ import gc
|
||||
from benchmark_utils import TimeCollector
|
||||
from tabulate import tabulate
|
||||
|
||||
from vllm.utils import FlexibleArgumentParser
|
||||
from vllm.utils.argparse_utils import FlexibleArgumentParser
|
||||
from vllm.v1.core.block_pool import BlockPool
|
||||
|
||||
|
||||
|
||||
@@ -46,7 +46,7 @@ import time
|
||||
|
||||
from vllm import LLM, SamplingParams
|
||||
from vllm.engine.arg_utils import EngineArgs
|
||||
from vllm.utils import FlexibleArgumentParser
|
||||
from vllm.utils.argparse_utils import FlexibleArgumentParser
|
||||
|
||||
|
||||
def test_long_document_qa(llm=None, sampling_params=None, prompts=None):
|
||||
|
||||
@@ -19,7 +19,7 @@ from vllm.config import (
|
||||
VllmConfig,
|
||||
)
|
||||
from vllm.platforms import current_platform
|
||||
from vllm.utils import FlexibleArgumentParser
|
||||
from vllm.utils.argparse_utils import FlexibleArgumentParser
|
||||
from vllm.v1.spec_decode.ngram_proposer import NgramProposer
|
||||
from vllm.v1.worker.gpu_input_batch import InputBatch
|
||||
from vllm.v1.worker.gpu_model_runner import GPUModelRunner
|
||||
|
||||
@@ -37,7 +37,7 @@ from transformers import PreTrainedTokenizerBase
|
||||
|
||||
from vllm import LLM, SamplingParams
|
||||
from vllm.engine.arg_utils import EngineArgs
|
||||
from vllm.utils import FlexibleArgumentParser
|
||||
from vllm.utils.argparse_utils import FlexibleArgumentParser
|
||||
|
||||
try:
|
||||
from vllm.transformers_utils.tokenizer import get_tokenizer
|
||||
|
||||
@@ -11,7 +11,7 @@ import time
|
||||
from transformers import AutoTokenizer, PreTrainedTokenizerBase
|
||||
|
||||
from vllm.engine.arg_utils import EngineArgs
|
||||
from vllm.utils import FlexibleArgumentParser
|
||||
from vllm.utils.argparse_utils import FlexibleArgumentParser
|
||||
|
||||
|
||||
# Select a equi-probable random priority
|
||||
|
||||
@@ -51,7 +51,7 @@ except ImportError:
|
||||
from backend_request_func import get_tokenizer
|
||||
|
||||
try:
|
||||
from vllm.utils import FlexibleArgumentParser
|
||||
from vllm.utils.argparse_utils import FlexibleArgumentParser
|
||||
except ImportError:
|
||||
from argparse import ArgumentParser as FlexibleArgumentParser
|
||||
|
||||
|
||||
@@ -15,7 +15,7 @@ from utils import make_rand_sparse_tensors
|
||||
from weight_shapes import WEIGHT_SHAPES
|
||||
|
||||
from vllm import _custom_ops as ops
|
||||
from vllm.utils import FlexibleArgumentParser
|
||||
from vllm.utils.argparse_utils import FlexibleArgumentParser
|
||||
|
||||
DEFAULT_MODELS = list(WEIGHT_SHAPES.keys())
|
||||
DEFAULT_BATCH_SIZES = [1, 16, 32, 64, 128, 256, 512]
|
||||
|
||||
@@ -18,7 +18,8 @@ from vllm import _custom_ops as ops
|
||||
from vllm.model_executor.layers.quantization.utils.fp8_utils import (
|
||||
w8a8_triton_block_scaled_mm,
|
||||
)
|
||||
from vllm.utils import FlexibleArgumentParser, cdiv
|
||||
from vllm.utils.argparse_utils import FlexibleArgumentParser
|
||||
from vllm.utils.math_utils import cdiv
|
||||
|
||||
DEFAULT_MODELS = list(WEIGHT_SHAPES.keys())
|
||||
DEFAULT_BATCH_SIZES = [1, 16, 32, 64, 128, 256, 512]
|
||||
|
||||
@@ -10,7 +10,7 @@ import torch
|
||||
from vllm.model_executor.layers.quantization.input_quant_fp8 import QuantFP8
|
||||
from vllm.model_executor.layers.quantization.utils.quant_utils import GroupShape
|
||||
from vllm.triton_utils import triton
|
||||
from vllm.utils import FlexibleArgumentParser
|
||||
from vllm.utils.argparse_utils import FlexibleArgumentParser
|
||||
from vllm.utils.torch_utils import STR_DTYPE_TO_TORCH_DTYPE
|
||||
|
||||
|
||||
|
||||
@@ -10,7 +10,7 @@ import vllm.model_executor.layers.activation # noqa F401
|
||||
from vllm.model_executor.custom_op import CustomOp
|
||||
from vllm.platforms import current_platform
|
||||
from vllm.triton_utils import triton
|
||||
from vllm.utils import FlexibleArgumentParser
|
||||
from vllm.utils.argparse_utils import FlexibleArgumentParser
|
||||
from vllm.utils.torch_utils import STR_DTYPE_TO_TORCH_DTYPE
|
||||
|
||||
batch_size_range = [1, 16, 32, 64, 128]
|
||||
|
||||
@@ -28,7 +28,7 @@ except ImportError as e:
|
||||
|
||||
from bitblas import Matmul, MatmulConfig, auto_detect_nvidia_target
|
||||
|
||||
from vllm.utils import FlexibleArgumentParser
|
||||
from vllm.utils.argparse_utils import FlexibleArgumentParser
|
||||
|
||||
parser = FlexibleArgumentParser(
|
||||
description="Benchmark BitBLAS int4 on a specific target."
|
||||
|
||||
@@ -20,7 +20,7 @@ from vllm.model_executor.layers.fused_moe.config import (
|
||||
from vllm.model_executor.layers.fused_moe.cutlass_moe import cutlass_moe_fp4
|
||||
from vllm.model_executor.layers.fused_moe.fused_moe import fused_experts, fused_topk
|
||||
from vllm.scalar_type import scalar_types
|
||||
from vllm.utils import FlexibleArgumentParser
|
||||
from vllm.utils.argparse_utils import FlexibleArgumentParser
|
||||
|
||||
WEIGHT_SHAPES_MOE = {
|
||||
"nvidia/DeepSeek-R1-FP4": [
|
||||
|
||||
@@ -14,7 +14,7 @@ from vllm.model_executor.layers.fused_moe.config import fp8_w8a8_moe_quant_confi
|
||||
from vllm.model_executor.layers.fused_moe.cutlass_moe import cutlass_moe_fp8
|
||||
from vllm.model_executor.layers.fused_moe.fused_moe import fused_experts, fused_topk
|
||||
from vllm.platforms import current_platform
|
||||
from vllm.utils import FlexibleArgumentParser
|
||||
from vllm.utils.argparse_utils import FlexibleArgumentParser
|
||||
|
||||
# Weight shapes for different models: [num_experts, topk, hidden_size,
|
||||
# intermediate_size]
|
||||
|
||||
@@ -39,7 +39,7 @@ from vllm.distributed.device_communicators.pynccl_allocator import (
|
||||
)
|
||||
from vllm.distributed.device_communicators.symm_mem import SymmMemCommunicator
|
||||
from vllm.logger import init_logger
|
||||
from vllm.utils import FlexibleArgumentParser
|
||||
from vllm.utils.argparse_utils import FlexibleArgumentParser
|
||||
|
||||
logger = init_logger(__name__)
|
||||
|
||||
|
||||
@@ -13,7 +13,7 @@ from vllm.model_executor.layers.fused_moe.fused_moe import (
|
||||
fused_experts,
|
||||
fused_topk,
|
||||
)
|
||||
from vllm.utils import FlexibleArgumentParser
|
||||
from vllm.utils.argparse_utils import FlexibleArgumentParser
|
||||
|
||||
DEFAULT_MODELS = [
|
||||
"nm-testing/Mixtral-8x7B-Instruct-v0.1",
|
||||
|
||||
@@ -7,7 +7,7 @@ import torch
|
||||
|
||||
from vllm.model_executor.layers.layernorm import RMSNorm
|
||||
from vllm.platforms import current_platform
|
||||
from vllm.utils import FlexibleArgumentParser
|
||||
from vllm.utils.argparse_utils import FlexibleArgumentParser
|
||||
from vllm.utils.torch_utils import STR_DTYPE_TO_TORCH_DTYPE
|
||||
|
||||
|
||||
|
||||
@@ -25,7 +25,7 @@ if HAS_TRITON:
|
||||
from vllm.lora.ops.triton_ops import LoRAKernelMeta, lora_expand, lora_shrink
|
||||
from vllm.lora.ops.triton_ops.utils import _LORA_A_PTR_DICT, _LORA_B_PTR_DICT
|
||||
|
||||
from vllm.utils import FlexibleArgumentParser
|
||||
from vllm.utils.argparse_utils import FlexibleArgumentParser
|
||||
|
||||
DEFAULT_MODELS = list(WEIGHT_SHAPES.keys())
|
||||
DEFAULT_TP_SIZES = [1]
|
||||
|
||||
@@ -33,7 +33,7 @@ from vllm.model_executor.layers.quantization.utils.quant_utils import (
|
||||
quantize_weights,
|
||||
)
|
||||
from vllm.scalar_type import ScalarType, scalar_types
|
||||
from vllm.utils import FlexibleArgumentParser
|
||||
from vllm.utils.argparse_utils import FlexibleArgumentParser
|
||||
|
||||
DEFAULT_MODELS = ["meta-llama/Llama-3-8b", "meta-llama/Llama-2-70b-hf"]
|
||||
DEFAULT_BATCH_SIZES = [1, 16, 32, 64, 128, 256, 512, 1024]
|
||||
|
||||
@@ -44,7 +44,7 @@ from vllm.model_executor.layers.quantization.utils.quant_utils import (
|
||||
sort_weights,
|
||||
)
|
||||
from vllm.scalar_type import ScalarType, scalar_types
|
||||
from vllm.utils import FlexibleArgumentParser
|
||||
from vllm.utils.argparse_utils import FlexibleArgumentParser
|
||||
|
||||
DEFAULT_MODELS = ["meta-llama/Llama-2-7b-hf/TP1"]
|
||||
DEFAULT_BATCH_SIZES = [1, 16, 32, 64, 128, 256, 512, 1024, 2048, 4096, 8192]
|
||||
|
||||
@@ -22,7 +22,7 @@ from vllm.model_executor.layers.fused_moe.fused_moe import *
|
||||
from vllm.platforms import current_platform
|
||||
from vllm.transformers_utils.config import get_config
|
||||
from vllm.triton_utils import triton
|
||||
from vllm.utils import FlexibleArgumentParser
|
||||
from vllm.utils.argparse_utils import FlexibleArgumentParser
|
||||
|
||||
FP8_DTYPE = current_platform.fp8_dtype()
|
||||
|
||||
|
||||
@@ -17,7 +17,7 @@ from vllm.model_executor.layers.fused_moe.moe_permute_unpermute import (
|
||||
)
|
||||
from vllm.model_executor.layers.fused_moe.utils import _fp8_quantize
|
||||
from vllm.platforms import current_platform
|
||||
from vllm.utils import FlexibleArgumentParser
|
||||
from vllm.utils.argparse_utils import FlexibleArgumentParser
|
||||
|
||||
FP8_DTYPE = current_platform.fp8_dtype()
|
||||
|
||||
|
||||
@@ -39,7 +39,7 @@ import torch
|
||||
from vllm.model_executor.layers.rotary_embedding import get_rope
|
||||
from vllm.platforms import current_platform
|
||||
from vllm.transformers_utils.config import get_config
|
||||
from vllm.utils import FlexibleArgumentParser
|
||||
from vllm.utils.argparse_utils import FlexibleArgumentParser
|
||||
|
||||
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
||||
|
||||
|
||||
@@ -9,7 +9,7 @@ import torch
|
||||
from vllm import _custom_ops as ops
|
||||
from vllm.logger import init_logger
|
||||
from vllm.platforms import current_platform
|
||||
from vllm.utils import FlexibleArgumentParser
|
||||
from vllm.utils.argparse_utils import FlexibleArgumentParser
|
||||
from vllm.utils.torch_utils import (
|
||||
STR_DTYPE_TO_TORCH_DTYPE,
|
||||
create_kv_caches_with_random,
|
||||
|
||||
@@ -7,7 +7,7 @@ import torch
|
||||
|
||||
from vllm import _custom_ops as ops
|
||||
from vllm.platforms import current_platform
|
||||
from vllm.utils import FlexibleArgumentParser
|
||||
from vllm.utils.argparse_utils import FlexibleArgumentParser
|
||||
from vllm.utils.torch_utils import STR_DTYPE_TO_TORCH_DTYPE
|
||||
|
||||
|
||||
|
||||
@@ -9,7 +9,7 @@ from tabulate import tabulate
|
||||
from vllm import _custom_ops as ops
|
||||
from vllm.logger import init_logger
|
||||
from vllm.platforms import current_platform
|
||||
from vllm.utils import FlexibleArgumentParser
|
||||
from vllm.utils.argparse_utils import FlexibleArgumentParser
|
||||
from vllm.utils.torch_utils import (
|
||||
STR_DTYPE_TO_TORCH_DTYPE,
|
||||
create_kv_caches_with_random,
|
||||
|
||||
@@ -12,7 +12,7 @@ from vllm.attention.ops.triton_reshape_and_cache_flash import (
|
||||
)
|
||||
from vllm.logger import init_logger
|
||||
from vllm.platforms import current_platform
|
||||
from vllm.utils import FlexibleArgumentParser
|
||||
from vllm.utils.argparse_utils import FlexibleArgumentParser
|
||||
from vllm.utils.torch_utils import (
|
||||
STR_DTYPE_TO_TORCH_DTYPE,
|
||||
create_kv_caches_with_random_flash,
|
||||
|
||||
@@ -8,7 +8,7 @@ import torch
|
||||
|
||||
from vllm.model_executor.layers.rotary_embedding import RotaryEmbedding, get_rope
|
||||
from vllm.platforms import current_platform
|
||||
from vllm.utils import FlexibleArgumentParser
|
||||
from vllm.utils.argparse_utils import FlexibleArgumentParser
|
||||
|
||||
|
||||
def benchmark_rope_kernels_multi_lora(
|
||||
|
||||
@@ -8,7 +8,7 @@ from datetime import datetime
|
||||
import flashinfer
|
||||
import torch
|
||||
|
||||
from vllm.utils import round_up
|
||||
from vllm.utils.math_utils import round_up
|
||||
|
||||
FLOAT32_BYTES = torch.finfo(torch.float).bits // 8
|
||||
FP8_DTYPE = torch.float8_e4m3fn
|
||||
|
||||
@@ -8,7 +8,7 @@ from datetime import datetime
|
||||
import flashinfer
|
||||
import torch
|
||||
|
||||
from vllm.utils import round_up
|
||||
from vllm.utils.math_utils import round_up
|
||||
|
||||
FLOAT32_BYTES = torch.finfo(torch.float).bits // 8
|
||||
FP8_DTYPE = torch.float8_e4m3fn
|
||||
|
||||
@@ -18,7 +18,7 @@ from vllm.model_executor.layers.quantization.utils.fp8_utils import (
|
||||
)
|
||||
from vllm.platforms import current_platform
|
||||
from vllm.triton_utils import triton
|
||||
from vllm.utils import FlexibleArgumentParser
|
||||
from vllm.utils.argparse_utils import FlexibleArgumentParser
|
||||
|
||||
mp.set_start_method("spawn", force=True)
|
||||
|
||||
|
||||
@@ -11,7 +11,7 @@ import regex as re
|
||||
import seaborn as sns
|
||||
from torch.utils.benchmark import Measurement as TMeasurement
|
||||
|
||||
from vllm.utils import FlexibleArgumentParser
|
||||
from vllm.utils.argparse_utils import FlexibleArgumentParser
|
||||
|
||||
if __name__ == "__main__":
|
||||
parser = FlexibleArgumentParser(
|
||||
|
||||
@@ -5,7 +5,7 @@ import cProfile
|
||||
import pstats
|
||||
|
||||
from vllm import LLM, SamplingParams
|
||||
from vllm.utils import FlexibleArgumentParser
|
||||
from vllm.utils.argparse_utils import FlexibleArgumentParser
|
||||
|
||||
# A very long prompt, total number of tokens is about 15k.
|
||||
LONG_PROMPT = ["You are an expert in large language models, aren't you?"] * 1000
|
||||
|
||||
Reference in New Issue
Block a user