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

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

View File

@@ -55,12 +55,9 @@ class OvisTEModel(sd1_clip.SD1ClipModel):
return out, pooled, {}
def te(dtype_llama=None, llama_scaled_fp8=None, llama_quantization_metadata=None):
def te(dtype_llama=None, llama_quantization_metadata=None):
class OvisTEModel_(OvisTEModel):
def __init__(self, device="cpu", dtype=None, model_options={}):
if llama_scaled_fp8 is not None and "scaled_fp8" not in model_options:
model_options = model_options.copy()
model_options["scaled_fp8"] = llama_scaled_fp8
if dtype_llama is not None:
dtype = dtype_llama
if llama_quantization_metadata is not None: