[Quantization] Migrate GGUF diffusion model support to OOT plugin (#4769)

Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
Signed-off-by: Isotr0py <Isotr0py@outlook.com>
This commit is contained in:
Isotr0py authored and GitHub committed 2026-07-11 18:48:43 +00:00
1 parent ff6f0da271
commit b95905deb1
35 files changed
+272 -1614

No files matched your search

@@ -280,7 +280,6 @@ def _build_quantization_config(
quantization: str | None,
quantization_config_json: str | None,
ignored_layers: str | None,
gguf_model: str | None,
) -> dict[str, Any] | None:
explicit = _load_json_arg(quantization_config_json)
if explicit is not None:
@@ -292,8 +291,6 @@ def _build_quantization_config(
ignored = _split_csv(ignored_layers)
if ignored:
config["ignored_layers"] = ignored
if gguf_model is not None:
config["gguf_model"] = gguf_model
return config
@@ -305,7 +302,6 @@ def _build_variant_config(args: argparse.Namespace, variant: str) -> VariantConf
quantization=quantization,
quantization_config_json=getattr(args, f"{variant}_quantization_config_json"),
ignored_layers=ignored_layers,
gguf_model=getattr(args, f"{variant}_gguf_model"),
)
return VariantConfig(
label=variant,
@@ -554,8 +550,6 @@ def parse_args() -> argparse.Namespace:
parser.add_argument("--reference-ignored-layers", help="Comma-separated reference ignored layer patterns.")
parser.add_argument("--candidate-ignored-layers", help="Comma-separated candidate ignored layer patterns.")
parser.add_argument("--ignored-layers", help="Alias for --candidate-ignored-layers.")
parser.add_argument("--reference-gguf-model", help="Reference GGUF file or HF reference when using gguf.")
parser.add_argument("--candidate-gguf-model", help="Candidate GGUF file or HF reference when using gguf.")
parser.add_argument("--reference-load-format", default="default", help="Reference diffusion load format.")
parser.add_argument("--candidate-load-format", default="default", help="Candidate diffusion load format.")
parser.add_argument("--prompt", default="a cup of coffee on the table")