[Misc] Move print_*_once from utils to logger (#11298)
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk> Signed-off-by: Maxime Fournioux <55544262+mfournioux@users.noreply.github.com> Co-authored-by: Maxime Fournioux <55544262+mfournioux@users.noreply.github.com>
This commit is contained in:
co-authored by
Maxime Fournioux
parent
730e9592e9
commit
d848800e88
@@ -5,7 +5,6 @@ import torch.nn as nn
|
||||
from vllm.config import get_current_vllm_config
|
||||
from vllm.logger import init_logger
|
||||
from vllm.platforms import current_platform
|
||||
from vllm.utils import print_warning_once
|
||||
|
||||
logger = init_logger(__name__)
|
||||
|
||||
@@ -91,7 +90,7 @@ class CustomOp(nn.Module):
|
||||
compilation_config = get_current_vllm_config().compilation_config
|
||||
custom_ops = compilation_config.custom_ops
|
||||
if not hasattr(cls, "name"):
|
||||
print_warning_once(
|
||||
logger.warning_once(
|
||||
f"Custom op {cls.__name__} was not registered, "
|
||||
f"which means it won't appear in the op registry. "
|
||||
f"It will be enabled/disabled based on the global settings.")
|
||||
|
||||
+5
-3
@@ -8,6 +8,7 @@ from compressed_tensors.quantization import QuantizationStrategy
|
||||
|
||||
import vllm.model_executor.layers.fused_moe # noqa
|
||||
from vllm import _custom_ops as ops
|
||||
from vllm.logger import init_logger
|
||||
from vllm.model_executor.layers.fused_moe import (FusedMoE, FusedMoEMethodBase,
|
||||
FusedMoeWeightScaleSupported)
|
||||
from vllm.model_executor.layers.quantization.compressed_tensors.schemes import (
|
||||
@@ -16,7 +17,8 @@ from vllm.model_executor.layers.quantization.utils.w8a8_utils import (
|
||||
all_close_1d, normalize_e4m3fn_to_e4m3fnuz, per_tensor_dequantize)
|
||||
from vllm.model_executor.utils import set_weight_attrs
|
||||
from vllm.platforms import current_platform
|
||||
from vllm.utils import print_warning_once
|
||||
|
||||
logger = init_logger(__name__)
|
||||
|
||||
|
||||
class GPTQMarlinState(Enum):
|
||||
@@ -142,10 +144,10 @@ class CompressedTensorsW8A8Fp8MoEMethod(CompressedTensorsMoEMethod):
|
||||
"activation scales are None.")
|
||||
if (not all_close_1d(layer.w13_input_scale)
|
||||
or not all_close_1d(layer.w2_input_scale)):
|
||||
print_warning_once(
|
||||
logger.warning_once(
|
||||
"Found input_scales that are not equal for "
|
||||
"fp8 MoE layer. Using the maximum across experts "
|
||||
"for each layer. ")
|
||||
"for each layer.")
|
||||
layer.w13_input_scale = torch.nn.Parameter(
|
||||
layer.w13_input_scale.max(), requires_grad=False)
|
||||
layer.w2_input_scale = torch.nn.Parameter(
|
||||
|
||||
@@ -28,7 +28,6 @@ from vllm.model_executor.parameter import (BlockQuantScaleParameter,
|
||||
PerTensorScaleParameter)
|
||||
from vllm.model_executor.utils import set_weight_attrs
|
||||
from vllm.platforms import current_platform
|
||||
from vllm.utils import print_warning_once
|
||||
|
||||
ACTIVATION_SCHEMES = ["static", "dynamic"]
|
||||
|
||||
@@ -539,10 +538,10 @@ class Fp8MoEMethod(FusedMoEMethodBase):
|
||||
"activation scales are None.")
|
||||
if (not all_close_1d(layer.w13_input_scale)
|
||||
or not all_close_1d(layer.w2_input_scale)):
|
||||
print_warning_once(
|
||||
logger.warning_once(
|
||||
"Found input_scales that are not equal for "
|
||||
"fp8 MoE layer. Using the maximum across experts "
|
||||
"for each layer. ")
|
||||
"for each layer.")
|
||||
layer.w13_input_scale = torch.nn.Parameter(
|
||||
layer.w13_input_scale.max(), requires_grad=False)
|
||||
layer.w2_input_scale = torch.nn.Parameter(
|
||||
|
||||
@@ -1,8 +1,10 @@
|
||||
import torch
|
||||
|
||||
from vllm.logger import init_logger
|
||||
from vllm.model_executor.layers.quantization.base_config import (
|
||||
QuantizationConfig, QuantizeMethodBase)
|
||||
from vllm.utils import print_warning_once
|
||||
|
||||
logger = init_logger(__name__)
|
||||
|
||||
|
||||
class BaseKVCacheMethod(QuantizeMethodBase):
|
||||
@@ -67,7 +69,7 @@ class BaseKVCacheMethod(QuantizeMethodBase):
|
||||
layer._v_scale = v_scale
|
||||
if (layer._k_scale == 1.0 and layer._v_scale == 1.0
|
||||
and "e5m2" not in layer.kv_cache_dtype):
|
||||
print_warning_once(
|
||||
logger.warning_once(
|
||||
"Using KV cache scaling factor 1.0 for fp8_e4m3. This "
|
||||
"may cause accuracy issues. Please make sure k/v_scale "
|
||||
"scaling factors are available in the fp8 checkpoint.")
|
||||
|
||||
@@ -3,11 +3,13 @@ from typing import Optional
|
||||
import torch
|
||||
|
||||
import vllm._custom_ops as ops
|
||||
from vllm.logger import init_logger
|
||||
from vllm.platforms import current_platform
|
||||
from vllm.utils import print_warning_once
|
||||
|
||||
from .marlin_utils import marlin_make_workspace, marlin_permute_scales
|
||||
|
||||
logger = init_logger(__name__)
|
||||
|
||||
|
||||
def is_fp8_marlin_supported():
|
||||
return current_platform.has_device_capability(80)
|
||||
@@ -47,7 +49,7 @@ def apply_fp8_marlin_linear(
|
||||
|
||||
def prepare_fp8_layer_for_marlin(layer: torch.nn.Module,
|
||||
strategy: str = "tensor") -> None:
|
||||
print_warning_once(
|
||||
logger.warning_once(
|
||||
"Your GPU does not have native support for FP8 computation but "
|
||||
"FP8 quantization is being used. Weight-only FP8 compression will "
|
||||
"be used leveraging the Marlin kernel. This may degrade "
|
||||
|
||||
@@ -25,7 +25,7 @@ from vllm.model_executor.layers.quantization import (QuantizationConfig,
|
||||
get_quantization_config)
|
||||
from vllm.model_executor.layers.quantization.schema import QuantParamSchema
|
||||
from vllm.platforms import current_platform
|
||||
from vllm.utils import PlaceholderModule, print_warning_once
|
||||
from vllm.utils import PlaceholderModule
|
||||
|
||||
try:
|
||||
from runai_model_streamer import SafetensorsStreamer
|
||||
@@ -673,7 +673,7 @@ def maybe_remap_kv_scale_name(name: str, params_dict: dict) -> Optional[str]:
|
||||
None: If the remapped name is not found in params_dict.
|
||||
"""
|
||||
if name.endswith(".kv_scale"):
|
||||
print_warning_once(
|
||||
logger.warning_once(
|
||||
"DEPRECATED. Found kv_scale in the checkpoint. "
|
||||
"This format is deprecated in favor of separate k_scale and "
|
||||
"v_scale tensors and will be removed in a future release. "
|
||||
@@ -682,7 +682,7 @@ def maybe_remap_kv_scale_name(name: str, params_dict: dict) -> Optional[str]:
|
||||
# NOTE: we remap the deprecated kv_scale to k_scale
|
||||
remapped_name = name.replace(".kv_scale", ".attn.k_scale")
|
||||
if remapped_name not in params_dict:
|
||||
print_warning_once(
|
||||
logger.warning_once(
|
||||
f"Found kv_scale in the checkpoint (e.g. {name}), "
|
||||
"but not found the expected name in the model "
|
||||
f"(e.g. {remapped_name}). kv_scale is "
|
||||
@@ -695,7 +695,7 @@ def maybe_remap_kv_scale_name(name: str, params_dict: dict) -> Optional[str]:
|
||||
if name.endswith(scale_name):
|
||||
remapped_name = name.replace(scale_name, f".attn{scale_name}")
|
||||
if remapped_name not in params_dict:
|
||||
print_warning_once(
|
||||
logger.warning_once(
|
||||
f"Found {scale_name} in the checkpoint (e.g. {name}), "
|
||||
"but not found the expected name in the model "
|
||||
f"(e.g. {remapped_name}). {scale_name} is "
|
||||
|
||||
@@ -11,6 +11,7 @@ from transformers import (BatchFeature, ChameleonConfig, ChameleonProcessor,
|
||||
from vllm.attention import Attention, AttentionMetadata
|
||||
from vllm.config import CacheConfig, VllmConfig
|
||||
from vllm.distributed import get_pp_group, get_tensor_model_parallel_world_size
|
||||
from vllm.logger import init_logger
|
||||
from vllm.model_executor.layers.activation import SiluAndMul
|
||||
from vllm.model_executor.layers.layernorm import RMSNorm
|
||||
from vllm.model_executor.layers.linear import (MergedColumnParallelLinear,
|
||||
@@ -35,13 +36,14 @@ from vllm.multimodal.processing import (BaseMultiModalProcessor,
|
||||
BaseProcessingInfo, PromptReplacement)
|
||||
from vllm.multimodal.profiling import BaseDummyInputsBuilder, ProcessorInputs
|
||||
from vllm.sequence import IntermediateTensors
|
||||
from vllm.utils import print_warning_once
|
||||
|
||||
from .interfaces import SupportsMultiModal, SupportsPP
|
||||
from .utils import (is_pp_missing_parameter,
|
||||
make_empty_intermediate_tensors_factory, make_layers,
|
||||
maybe_prefix, merge_multimodal_embeddings)
|
||||
|
||||
logger = init_logger(__name__)
|
||||
|
||||
|
||||
class ChameleonImagePixelInputs(TypedDict):
|
||||
type: Literal["pixel_values"]
|
||||
@@ -1111,7 +1113,7 @@ class ChameleonForConditionalGeneration(nn.Module, SupportsMultiModal,
|
||||
remapped_kv_scale_name = name.replace(
|
||||
".kv_scale", ".attn.kv_scale")
|
||||
if remapped_kv_scale_name not in params_dict:
|
||||
print_warning_once(
|
||||
logger.warning_once(
|
||||
"Found kv scale in the checkpoint (e.g. "
|
||||
f"{name}), but not found the expected name in "
|
||||
f"the model (e.g. {remapped_kv_scale_name}). "
|
||||
|
||||
@@ -20,6 +20,7 @@ from vllm.attention import Attention, AttentionMetadata
|
||||
from vllm.compilation.decorators import support_torch_compile
|
||||
from vllm.config import CacheConfig, VllmConfig
|
||||
from vllm.distributed import get_pp_group, get_tensor_model_parallel_world_size
|
||||
from vllm.logger import init_logger
|
||||
from vllm.model_executor.layers.fused_moe import FusedMoE
|
||||
from vllm.model_executor.layers.layernorm import RMSNorm
|
||||
from vllm.model_executor.layers.linear import (QKVParallelLinear,
|
||||
@@ -34,13 +35,14 @@ from vllm.model_executor.layers.vocab_parallel_embedding import (
|
||||
from vllm.model_executor.model_loader.weight_utils import default_weight_loader
|
||||
from vllm.model_executor.sampling_metadata import SamplingMetadata
|
||||
from vllm.sequence import IntermediateTensors
|
||||
from vllm.utils import print_warning_once
|
||||
|
||||
from .interfaces import SupportsPP
|
||||
from .utils import (is_pp_missing_parameter,
|
||||
make_empty_intermediate_tensors_factory, make_layers,
|
||||
maybe_prefix)
|
||||
|
||||
logger = init_logger(__name__)
|
||||
|
||||
|
||||
class OlmoeMoE(nn.Module):
|
||||
"""A tensor-parallel MoE implementation for Olmoe that shards each expert
|
||||
@@ -446,7 +448,7 @@ class OlmoeForCausalLM(nn.Module, SupportsPP):
|
||||
remapped_kv_scale_name = name.replace(
|
||||
".kv_scale", ".attn.kv_scale")
|
||||
if remapped_kv_scale_name not in params_dict:
|
||||
print_warning_once(
|
||||
logger.warning_once(
|
||||
"Found kv scale in the checkpoint "
|
||||
f"(e.g. {name}), but not found the expected "
|
||||
f"name in the model "
|
||||
|
||||
@@ -34,6 +34,7 @@ from vllm.config import CacheConfig, VllmConfig
|
||||
from vllm.distributed import (get_pp_group,
|
||||
get_tensor_model_parallel_world_size,
|
||||
tensor_model_parallel_all_reduce)
|
||||
from vllm.logger import init_logger
|
||||
from vllm.model_executor.layers.activation import SiluAndMul
|
||||
from vllm.model_executor.layers.fused_moe import FusedMoE
|
||||
from vllm.model_executor.layers.layernorm import RMSNorm
|
||||
@@ -50,13 +51,14 @@ from vllm.model_executor.layers.vocab_parallel_embedding import (
|
||||
from vllm.model_executor.model_loader.weight_utils import default_weight_loader
|
||||
from vllm.model_executor.sampling_metadata import SamplingMetadata
|
||||
from vllm.sequence import IntermediateTensors
|
||||
from vllm.utils import print_warning_once
|
||||
|
||||
from .interfaces import SupportsPP
|
||||
from .utils import (extract_layer_index, is_pp_missing_parameter,
|
||||
make_empty_intermediate_tensors_factory, make_layers,
|
||||
maybe_prefix)
|
||||
|
||||
logger = init_logger(__name__)
|
||||
|
||||
|
||||
class Qwen2MoeMLP(nn.Module):
|
||||
|
||||
@@ -524,7 +526,7 @@ class Qwen2MoeForCausalLM(nn.Module, SupportsPP):
|
||||
remapped_kv_scale_name = name.replace(
|
||||
".kv_scale", ".attn.kv_scale")
|
||||
if remapped_kv_scale_name not in params_dict:
|
||||
print_warning_once(
|
||||
logger.warning_once(
|
||||
"Found kv scale in the checkpoint "
|
||||
f"(e.g. {name}), but not found the expected "
|
||||
f"name in the model "
|
||||
|
||||
@@ -7,8 +7,10 @@ from transformers import PretrainedConfig
|
||||
import vllm.envs as envs
|
||||
from vllm.attention.selector import (backend_name_to_enum,
|
||||
get_global_forced_attn_backend)
|
||||
from vllm.logger import init_logger
|
||||
from vllm.platforms import _Backend, current_platform
|
||||
from vllm.utils import print_warning_once
|
||||
|
||||
logger = init_logger(__name__)
|
||||
|
||||
_C = TypeVar("_C", bound=PretrainedConfig)
|
||||
|
||||
@@ -87,7 +89,7 @@ def get_vit_attn_backend(support_fa: bool = False) -> _Backend:
|
||||
if is_flash_attn_2_available():
|
||||
selected_backend = _Backend.FLASH_ATTN
|
||||
else:
|
||||
print_warning_once(
|
||||
logger.warning_once(
|
||||
"Current `vllm-flash-attn` has a bug inside vision module, "
|
||||
"so we use xformers backend instead. You can run "
|
||||
"`pip install flash-attn` to use flash-attention backend.")
|
||||
|
||||
Reference in New Issue
Block a user