[Rebase] Rebase to vllm v0.24.0 (#4709)
This commit is contained in:
1 parent
9a105bf90d
commit
a560ed184d
60 files changed
+789
-203
No files matched your search
@@ -18,7 +18,6 @@ from vllm.config.load import LoadConfig
|
||||
from vllm.logger import init_logger
|
||||
from vllm.model_executor.layers.quantization.base_config import QuantizeMethodBase
|
||||
from vllm.model_executor.model_loader.weight_utils import (
|
||||
download_gguf,
|
||||
download_safetensors_index_file_from_hf,
|
||||
download_weights_from_hf,
|
||||
filter_duplicate_safetensors_files,
|
||||
@@ -42,6 +41,39 @@ from vllm_omni.diffusion.models.diffusers_adapter.pipeline_diffusers_adapter imp
|
||||
from vllm_omni.diffusion.offloader.module_collector import ModuleDiscovery
|
||||
from vllm_omni.diffusion.registry import initialize_model
|
||||
|
||||
|
||||
# download_gguf was removed from upstream vLLM (commit 6635279d8).
|
||||
# Inlined from the last upstream version before the GGUF plugin migration.
|
||||
def download_gguf(
|
||||
repo_id: str,
|
||||
quant_type: str,
|
||||
cache_dir: str | None = None,
|
||||
revision: str | None = None,
|
||||
ignore_patterns: str | list[str] | None = None,
|
||||
) -> str:
|
||||
allow_patterns = [
|
||||
f"*-{quant_type}.gguf",
|
||||
f"*-{quant_type}-*.gguf",
|
||||
f"*/*-{quant_type}.gguf",
|
||||
f"*/*-{quant_type}-*.gguf",
|
||||
]
|
||||
folder = download_weights_from_hf(
|
||||
model_name_or_path=repo_id,
|
||||
cache_dir=cache_dir,
|
||||
allow_patterns=allow_patterns,
|
||||
revision=revision,
|
||||
ignore_patterns=ignore_patterns,
|
||||
)
|
||||
local_files: list[str] = []
|
||||
for pattern in allow_patterns:
|
||||
glob_pattern = os.path.join(folder, pattern)
|
||||
local_files.extend(glob.glob(glob_pattern))
|
||||
if not local_files:
|
||||
raise ValueError(f"Downloaded GGUF files not found in {folder} for quant_type {quant_type}")
|
||||
local_files.sort(key=lambda x: (x.count("-"), x))
|
||||
return local_files[0]
|
||||
|
||||
|
||||
logger = init_logger(__name__)
|
||||
|
||||
|
||||
|
||||
@@ -6,10 +6,18 @@ from collections.abc import Generator
|
||||
from dataclasses import dataclass
|
||||
from typing import TYPE_CHECKING, Any
|
||||
|
||||
import gguf
|
||||
import numpy as np
|
||||
import torch
|
||||
|
||||
try:
|
||||
import gguf
|
||||
except ImportError:
|
||||
# Optional dependency: only required for GGUF-quantized diffusion models.
|
||||
# Keep this import lazy so a missing ``gguf`` does not break importing the
|
||||
# diffusion model loader (and thus inspection/loading of every other
|
||||
# diffusion model that imports through this package).
|
||||
gguf = None
|
||||
|
||||
if TYPE_CHECKING:
|
||||
from vllm_omni.diffusion.data import OmniDiffusionConfig
|
||||
from vllm_omni.diffusion.model_loader.diffusers_loader import (
|
||||
@@ -65,6 +73,11 @@ def gguf_quant_weights_iterator(gguf_file: str) -> Generator[tuple[str, torch.Te
|
||||
layer with different quant types.
|
||||
"""
|
||||
|
||||
if gguf is None:
|
||||
raise ImportError(
|
||||
"The 'gguf' package is required to load GGUF-quantized diffusion models. "
|
||||
"Install it with `pip install gguf`."
|
||||
)
|
||||
reader = gguf.GGUFReader(gguf_file)
|
||||
|
||||
for tensor in reader.tensors:
|
||||
|
||||
Reference in new issue
Block a user