Olive dev #2429
Replies: 4 comments 9 replies
-
|
i've made changes to i haven't made any changes to
def enable_olive_onchange():
from installer import installed, install, uninstall
if shared.opts.onnx_enable_olive:
if not installed('olive-ai', reload=True, quiet=True):
install('olive-ai', 'olive-ai')
else:
global is_available # pylint: disable=global-statement
is_available = False
if "olive" in sys.modules:
del sys.modules["olive"]
if shared.opts.diffusers_pipeline == 'ONNX Stable Diffusion with Olive':
shared.opts.diffusers_pipeline = 'ONNX Stable Diffusion'
if installed('olive-ai', reload=True, quiet=True):
uninstall('olive-ai')
def init_olive():
try:
if installer.opts['onnx_enable_olive']:
import olive.workflows # pylint: disable=unused-import
installer.log.debug('Load olive')
except Exception as e:
installer.log.error(f'Failed to load olive: {e}')
model_list += modelloader.load_diffusers_models(model_path=shared.opts.olive_sideloaded_models_path, command_path=shared.opts.olive_sideloaded_models_path, clear=False)
if shared.opts.olive_sideloaded_models_path in checkpoint_info.path:
try:
from modules.onnx import OnnxStableDiffusionPipeline
sd_model = OnnxStableDiffusionPipeline.from_pretrained(checkpoint_info.path, cache_dir=shared.opts.olive_sideloaded_models_path)
sd_model.model_type = sd_model.__class__.__name__
except Exception as e:
shared.log.error(f'Failed loading {op}: {checkpoint_info.path} olive={e}')
return |
Beta Was this translation helpful? Give feedback.
-
|
adding @Disty0 lets keep core required changes in dev (and refresh olive branch as needed with those changes) and olive branch modifications specific to olive, otherwise its going to be a nightmare to merge. some of changes in olive branch that impact stuff that i've mentioned in the first post should really be done in core directly and carefully. goal is that core should support olive, but olive should not make one-way changes to core. |
Beta Was this translation helpful? Give feedback.
-
|
Made the changes in dev (b3ef210) and merged dev to olive branch (1f57fb3...04e3370) |
Beta Was this translation helpful? Give feedback.
-
|
@lshqqytiger So I tried with another model but I lift the same error. When it tries to convert any model including v1-5-pruned-emaonly, I get multiple error with longer logs like this one: And If I retry: |
Beta Was this translation helpful? Give feedback.
-
|
Maybe you downgraded torch without activating virtual environment. |
Beta Was this translation helpful? Give feedback.
-
|
Right. Totally forgot. |
Beta Was this translation helpful? Give feedback.
-
|
If you don't have |
Beta Was this translation helpful? Give feedback.
-
|
Sorry I didn't answer. I did manage to have it working and converting my models. Now I wanted to know if it was normal that the second pass doesn't work (hires fix): This error shows up as long as the upscaler isn't set to Latent. |
Beta Was this translation helpful? Give feedback.
-
|
Second pass is not tested. I will have a look. |
Beta Was this translation helpful? Give feedback.
-
|
22:41:59-141796 ERROR Diffusers failed loading: |
Beta Was this translation helpful? Give feedback.
-
|
when trying to use sdxl |
Beta Was this translation helpful? Give feedback.
-
this thread is for olive dev only, do not post unless you're developing or testing experimental olive branch
Beta Was this translation helpful? Give feedback.
All reactions