jaidaken/ComfyUI
1
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

Make old scaled fp8 format use the new mixed quant ops system. (#11000)

Browse Source
This commit is contained in:
comfyanonymous
2025-12-05 11:35:42 -08:00
committed by GitHub
parent 0ec05b1481
commit 43071e3de3
24 changed files with 278 additions and 275 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
comfy/text_encoders/omnigen2.py
View File

@@ -32,12 +32,12 @@ class Omnigen2Model(sd1_clip.SD1ClipModel):
super().__init__(device=device, dtype=dtype, name="qwen25_3b", clip_model=Qwen25_3BModel, model_options=model_options)
def te(dtype_llama=None, llama_scaled_fp8=None):
def te(dtype_llama=None, llama_quantization_metadata=None):
class Omnigen2TEModel_(Omnigen2Model):
def __init__(self, device="cpu", dtype=None, model_options={}):
if llama_scaled_fp8 is not None and "scaled_fp8" not in model_options:
if llama_quantization_metadata is not None:
model_options = model_options.copy()
model_options["scaled_fp8"] = llama_scaled_fp8
model_options["quantization_metadata"] = llama_quantization_metadata
if dtype_llama is not None:
dtype = dtype_llama
super().__init__(device=device, dtype=dtype, model_options=model_options)