AI/diffusers
1
0
Fork 0
mirror of https://github.com/huggingface/diffusers.git synced 2026-01-27 17:22:53 +03:00
Code Activity

make parallel loading flag a part of constants. (#12137)

Browse Source
This commit is contained in:
Sayak Paul
2025-08-14 09:36:47 +05:30
committed by GitHub
parent 8c48ec05ed
commit 123506ee59
3 changed files with 4 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
src/diffusers/models/modeling_utils.py
View File

@@ -42,9 +42,8 @@ from ..quantizers import DiffusersAutoQuantizer, DiffusersQuantizer
from ..quantizers.quantization_config import QuantizationMethod
from ..utils import (
CONFIG_NAME,
ENV_VARS_TRUE_VALUES,
FLAX_WEIGHTS_NAME,
HF_PARALLEL_LOADING_FLAG,
HF_ENABLE_PARALLEL_LOADING,
SAFE_WEIGHTS_INDEX_NAME,
SAFETENSORS_WEIGHTS_NAME,
WEIGHTS_INDEX_NAME,
@@ -962,7 +961,7 @@ class ModelMixin(torch.nn.Module, PushToHubMixin):
dduf_entries: Optional[Dict[str, DDUFEntry]] = kwargs.pop("dduf_entries", None)
disable_mmap = kwargs.pop("disable_mmap", False)
is_parallel_loading_enabled = os.environ.get(HF_PARALLEL_LOADING_FLAG, "").upper() in ENV_VARS_TRUE_VALUES
is_parallel_loading_enabled = HF_ENABLE_PARALLEL_LOADING
if is_parallel_loading_enabled and not low_cpu_mem_usage:
raise NotImplementedError("Parallel loading is not supported when not using `low_cpu_mem_usage`.")

2
src/diffusers/utils/__init__.py
View File

@@ -25,8 +25,8 @@ from .constants import (
DIFFUSERS_DYNAMIC_MODULE_NAME,
FLAX_WEIGHTS_NAME,
GGUF_FILE_EXTENSION,
HF_ENABLE_PARALLEL_LOADING,
HF_MODULES_CACHE,
HF_PARALLEL_LOADING_FLAG,
HUGGINGFACE_CO_RESOLVE_ENDPOINT,
MIN_PEFT_VERSION,
ONNX_EXTERNAL_WEIGHTS_NAME,

2
src/diffusers/utils/constants.py
View File

@@ -44,7 +44,7 @@ DIFFUSERS_REQUEST_TIMEOUT = 60
DIFFUSERS_ATTN_BACKEND = os.getenv("DIFFUSERS_ATTN_BACKEND", "native")
DIFFUSERS_ATTN_CHECKS = os.getenv("DIFFUSERS_ATTN_CHECKS", "0") in ENV_VARS_TRUE_VALUES
DEFAULT_HF_PARALLEL_LOADING_WORKERS = 8
HF_PARALLEL_LOADING_FLAG = "HF_ENABLE_PARALLEL_LOADING"
HF_ENABLE_PARALLEL_LOADING = os.environ.get("HF_ENABLE_PARALLEL_LOADING", "").upper() in ENV_VARS_TRUE_VALUES
# Below should be `True` if the current version of `peft` and `transformers` are compatible with
# PEFT backend. Will automatically fall back to PEFT backend if the correct versions of the libraries are