[Feature] OTEL tracing during loading (#31162)

This commit is contained in:
emricksini-h
2026-02-05 16:59:28 -08:00
committed by GitHub
parent 91a07ff618
commit 325ab6b0a8
29 changed files with 873 additions and 280 deletions
@@ -14,6 +14,7 @@ from vllm.model_executor.model_loader.utils import (
process_weights_after_loading,
)
from vllm.platforms import current_platform
from vllm.tracing import instrument
from vllm.utils.mem_utils import format_gib
from vllm.utils.torch_utils import set_default_torch_dtype
@@ -37,6 +38,7 @@ class BaseModelLoader(ABC):
inplace weights loading for an already-initialized model"""
raise NotImplementedError
@instrument(span_name="Load model")
def load_model(
self, vllm_config: VllmConfig, model_config: ModelConfig, prefix: str = ""
) -> nn.Module:
@@ -30,6 +30,7 @@ from vllm.model_executor.model_loader.weight_utils import (
pt_weights_iterator,
safetensors_weights_iterator,
)
from vllm.tracing import instrument
from vllm.transformers_utils.repo_utils import list_filtered_repo_files
logger = init_logger(__name__)
@@ -274,6 +275,7 @@ class DefaultModelLoader(BaseModelLoader):
allow_patterns_overrides=None,
)
@instrument(span_name="Load weights")
def load_weights(self, model: nn.Module, model_config: ModelConfig) -> None:
if model_config.quantization == "torchao":
quant_config = get_quant_config(model_config, self.load_config)
@@ -23,11 +23,13 @@ from vllm.model_executor.model_loader.reload import (
set_torchao_reload_attrs,
)
from vllm.model_executor.models.interfaces import SupportsQuant
from vllm.tracing import instrument
from vllm.utils.platform_utils import is_pin_memory_available
logger = init_logger(__name__)
@instrument(span_name="Initialize model")
def initialize_model(
vllm_config: VllmConfig,
*,
@@ -36,6 +36,7 @@ from vllm.model_executor.layers.quantization import (
get_quantization_config,
)
from vllm.platforms import current_platform
from vllm.tracing import instrument
from vllm.utils.import_utils import PlaceholderModule
try:
@@ -443,6 +444,7 @@ def download_gguf(
return local_files[0]
@instrument(span_name="Download weights - HF")
def download_weights_from_hf(
model_name_or_path: str,
cache_dir: str | None,