mirror of https://github.com/vladmandic/automatic
90 lines
4.2 KiB
Python
90 lines
4.2 KiB
Python
import transformers
|
|
import diffusers
|
|
from modules import shared, devices, sd_models, model_quant, sd_hijack_te, sd_hijack_vae
|
|
|
|
|
|
def load_qwen(checkpoint_info, diffusers_load_config=None):
|
|
from pipelines import generic, qwen
|
|
if diffusers_load_config is None:
|
|
diffusers_load_config = {}
|
|
repo_id = sd_models.path_to_repo(checkpoint_info)
|
|
repo_subfolder = checkpoint_info.subfolder
|
|
sd_models.hf_auth_check(checkpoint_info)
|
|
transformer = None
|
|
|
|
load_args, _quant_args = model_quant.get_dit_args(diffusers_load_config, module='Model')
|
|
shared.log.debug(f'Load model: type=Qwen model="{checkpoint_info.name}" repo="{repo_id}" offload={shared.opts.diffusers_offload_mode} dtype={devices.dtype} args={load_args}')
|
|
|
|
if '2509' in repo_id :
|
|
cls_name = diffusers.QwenImageEditPlusPipeline
|
|
diffusers.pipelines.auto_pipeline.AUTO_TEXT2IMAGE_PIPELINES_MAPPING["qwen-image"] = diffusers.QwenImageEditPlusPipeline
|
|
diffusers.pipelines.auto_pipeline.AUTO_IMAGE2IMAGE_PIPELINES_MAPPING["qwen-image"] = diffusers.QwenImageEditPlusPipeline
|
|
diffusers.pipelines.auto_pipeline.AUTO_INPAINT_PIPELINES_MAPPING["qwen-image"] = diffusers.QwenImageEditPlusPipeline
|
|
elif 'Edit' in repo_id:
|
|
cls_name = diffusers.QwenImageEditPipeline
|
|
diffusers.pipelines.auto_pipeline.AUTO_TEXT2IMAGE_PIPELINES_MAPPING["qwen-image"] = diffusers.QwenImageEditPipeline
|
|
diffusers.pipelines.auto_pipeline.AUTO_IMAGE2IMAGE_PIPELINES_MAPPING["qwen-image"] = diffusers.QwenImageEditPipeline
|
|
diffusers.pipelines.auto_pipeline.AUTO_INPAINT_PIPELINES_MAPPING["qwen-image"] = diffusers.QwenImageEditPipeline
|
|
elif 'Layered' in repo_id:
|
|
cls_name = diffusers.QwenImageLayeredPipeline
|
|
diffusers.pipelines.auto_pipeline.AUTO_TEXT2IMAGE_PIPELINES_MAPPING["qwen-layered"] = diffusers.QwenImageLayeredPipeline
|
|
diffusers.pipelines.auto_pipeline.AUTO_IMAGE2IMAGE_PIPELINES_MAPPING["qwen-layered"] = diffusers.QwenImageLayeredPipeline
|
|
diffusers.pipelines.auto_pipeline.AUTO_INPAINT_PIPELINES_MAPPING["qwen-layered"] = diffusers.QwenImageLayeredPipeline
|
|
else:
|
|
cls_name = diffusers.QwenImagePipeline
|
|
diffusers.pipelines.auto_pipeline.AUTO_TEXT2IMAGE_PIPELINES_MAPPING["qwen-image"] = diffusers.QwenImagePipeline
|
|
diffusers.pipelines.auto_pipeline.AUTO_IMAGE2IMAGE_PIPELINES_MAPPING["qwen-image"] = diffusers.QwenImageImg2ImgPipeline
|
|
diffusers.pipelines.auto_pipeline.AUTO_INPAINT_PIPELINES_MAPPING["qwen-image"] = diffusers.QwenImageInpaintPipeline
|
|
|
|
if model_quant.check_nunchaku('Model'):
|
|
transformer = qwen.load_qwen_nunchaku(repo_id)
|
|
|
|
if 'Qwen-Image-Distill-Full' in repo_id:
|
|
repo_transformer = repo_id
|
|
transformer_subfolder = None
|
|
repo_id = 'Qwen/Qwen-Image'
|
|
else:
|
|
repo_transformer = repo_id
|
|
if repo_subfolder is not None:
|
|
transformer_subfolder = repo_subfolder + '/transformer'
|
|
else:
|
|
transformer_subfolder = "transformer"
|
|
|
|
if transformer is None:
|
|
transformer = generic.load_transformer(
|
|
repo_transformer,
|
|
subfolder=transformer_subfolder,
|
|
cls_name=diffusers.QwenImageTransformer2DModel,
|
|
load_config=diffusers_load_config,
|
|
modules_to_not_convert=["transformer_blocks.0.img_mod.1.weight"],
|
|
)
|
|
|
|
repo_te = 'Qwen/Qwen-Image'
|
|
text_encoder = generic.load_text_encoder(repo_te, cls_name=transformers.Qwen2_5_VLForConditionalGeneration, load_config=diffusers_load_config)
|
|
|
|
repo_id, repo_subfolder = qwen.check_qwen_pruning(repo_id, repo_subfolder)
|
|
pipe = cls_name.from_pretrained(
|
|
repo_id,
|
|
transformer=transformer,
|
|
text_encoder=text_encoder,
|
|
subfolder=repo_subfolder,
|
|
cache_dir=shared.opts.diffusers_dir,
|
|
**load_args,
|
|
)
|
|
pipe.task_args = {
|
|
'output_type': 'np',
|
|
}
|
|
if 'Layered' in repo_id:
|
|
pipe.task_args['use_en_prompt'] = True
|
|
pipe.task_args['cfg_normalize'] = False
|
|
pipe.task_args['layers'] = shared.opts.model_qwen_layers
|
|
pipe.task_args['resolution'] = 640
|
|
|
|
del text_encoder
|
|
del transformer
|
|
sd_hijack_te.init_hijack(pipe)
|
|
sd_hijack_vae.init_hijack(pipe)
|
|
|
|
devices.torch_gc()
|
|
return pipe
|