Report of some bugs that have had dealt with it or could not deal with.

[NaughtDZ]

File "J:\Dreambooth-Stable-Diffusion-cpu\main.py", line 18, in
from pytorch_lightning.utilities.distributed import rank_zero_only
ImportError: cannot import name 'rank_zero_only' from 'pytorch_lightning.utilities.distributed'

According to "AUTOMATIC1111/stable-diffusion-webui#4111"
The function was moved to a new script with a different name . Try to import it as :

from pytorch_lightning.utilities.rank_zero import rank_zero_only

And
File "J:\Dreambooth-Stable-Diffusion-cpu\ldm\models\diffusion\ddpm.py", line 21, in
Has same issue.

---

Traceback (most recent call last):
File "J:\Dreambooth-Stable-Diffusion-cpu\main.py", line 851, in
if trainer.global_rank == 0:
NameError: name 'trainer' is not defined. Did you mean: 'Trainer'?

I just change trainer to Traniner,and program can running,but dont know if have other issues.

---

Traceback (most recent call last):
File "J:\Dreambooth-Stable-Diffusion-cpu\main.py", line 670, in
trainer_kwargs["logger"] = instantiate_from_config(logger_cfg)
File "J:\Dreambooth-Stable-Diffusion-cpu\ldm\util.py", line 85, in instantiate_from_config
return get_obj_from_str(config["target"])(**config.get("params", dict()), **kwargs)
File "J:\Dreambooth-Stable-Diffusion-cpu\ldm\util.py", line 93, in get_obj_from_str
return getattr(importlib.import_module(module, package=None), cls)
AttributeError: module 'pytorch_lightning.loggers' has no attribute 'TestTubeLogger'. Did you mean: 'NeptuneLogger'?

thx:78Alpha

Replace all instances of TestTubeLogger with CSVLogger

[78Alpha]

Replace all instances of TestTubeLogger with CSVLogger

[78Alpha]

After altering all the necessary parts it will auto Kill after 1 epoch, so it seems to hit a wall.

Epoch 0: 0%| | 1/1010 [00:40<11:16:58, 40.26s/it, loss=0.625, v_num=0, train/loss_simple_step=0.604, train/loss_vlb_st Killed

[NaughtDZ]

New issue:
4.
Traceback (most recent call last):
File "J:\Dreambooth-Stable-Diffusion-cpu\main.py", line 830, in
trainer.fit(model, data)
File "C:\Users\Naught\AppData\Local\Programs\Python\Python310\lib\site-packages\pytorch_lightning\trainer\trainer.py", line 579, in fit
call._call_and_handle_interrupt(
File "C:\Users\Naught\AppData\Local\Programs\Python\Python310\lib\site-packages\pytorch_lightning\trainer\call.py", line 38, in _call_and_handle_interrupt
return trainer_fn(*args, **kwargs)
File "C:\Users\Naught\AppData\Local\Programs\Python\Python310\lib\site-packages\pytorch_lightning\trainer\trainer.py", line 621, in _fit_impl
self._run(model, ckpt_path=self.ckpt_path)
File "C:\Users\Naught\AppData\Local\Programs\Python\Python310\lib\site-packages\pytorch_lightning\trainer\trainer.py", line 984, in _run
verify_loop_configurations(self)
File "C:\Users\Naught\AppData\Local\Programs\Python\Python310\lib\site-packages\pytorch_lightning\trainer\configuration_validator.py", line 53, in verify_loop_configurations
_check_deprecated_callback_hooks(trainer)
File "C:\Users\Naught\AppData\Local\Programs\Python\Python310\lib\site-packages\pytorch_lightning\trainer\configuration_validator.py", line 266, in _check_deprecated_callback_hooks
raise RuntimeError(
RuntimeError: The Callback.on_pretrain_routine_start hook was removed in v1.8. Please use Callback.on_fit_start instead.

thx:78Alpha

Main.py, line 327 rename on_pretrain_routine_start to on_fit_start or on_fit_end

[78Alpha]

Main.py, line 327 or 330, rename on_pretrain_routine_start to on_fit_start or on_fit_end

[andreae293]

Regarding error.3
AttributeError: module 'pytorch_lightning.loggers' has no attribute 'TestTubeLogger'

did you guys install test-tube?
pip install test-tube

[NaughtDZ]

yes it working ,thx

[NaughtDZ]

New issue 5.

Epoch 0: 0%| | 0/2020 [00:00<?, ?it/s]Summoning checkpoint.
Traceback (most recent call last):
File "J:\Dreambooth-Stable-Diffusion-cpu\main.py", line 830, in
trainer.fit(model, data)
File "C:\Users\Naught\AppData\Local\Programs\Python\Python310\lib\site-packages\pytorch_lightning\trainer\trainer.py", line 579, in fit
call._call_and_handle_interrupt(
File "C:\Users\Naught\AppData\Local\Programs\Python\Python310\lib\site-packages\pytorch_lightning\trainer\call.py", line 38, in _call_and_handle_interrupt
return trainer_fn(*args, **kwargs)
File "C:\Users\Naught\AppData\Local\Programs\Python\Python310\lib\site-packages\pytorch_lightning\trainer\trainer.py", line 621, in _fit_impl
self._run(model, ckpt_path=self.ckpt_path)
File "C:\Users\Naught\AppData\Local\Programs\Python\Python310\lib\site-packages\pytorch_lightning\trainer\trainer.py", line 1058, in _run
results = self._run_stage()
File "C:\Users\Naught\AppData\Local\Programs\Python\Python310\lib\site-packages\pytorch_lightning\trainer\trainer.py", line 1137, in _run_stage
self._run_train()
File "C:\Users\Naught\AppData\Local\Programs\Python\Python310\lib\site-packages\pytorch_lightning\trainer\trainer.py", line 1160, in _run_train
self.fit_loop.run()
File "C:\Users\Naught\AppData\Local\Programs\Python\Python310\lib\site-packages\pytorch_lightning\loops\loop.py", line 199, in run
self.advance(*args, **kwargs)
File "C:\Users\Naught\AppData\Local\Programs\Python\Python310\lib\site-packages\pytorch_lightning\loops\fit_loop.py", line 267, in advance
self._outputs = self.epoch_loop.run(self._data_fetcher)
File "C:\Users\Naught\AppData\Local\Programs\Python\Python310\lib\site-packages\pytorch_lightning\loops\loop.py", line 199, in run
self.advance(*args, **kwargs)
File "C:\Users\Naught\AppData\Local\Programs\Python\Python310\lib\site-packages\pytorch_lightning\loops\epoch\training_epoch_loop.py", line 205, in advance
response = self.trainer._call_lightning_module_hook("on_train_batch_start", batch, batch_idx)
File "C:\Users\Naught\AppData\Local\Programs\Python\Python310\lib\site-packages\pytorch_lightning\trainer\trainer.py", line 1302, in _call_lightning_module_hook
output = fn(*args, **kwargs)
File "C:\Users\Naught\AppData\Local\Programs\Python\Python310\lib\site-packages\lightning_utilities\core\rank_zero.py", line 24, in wrapped_fn
return fn(*args, **kwargs)
File "C:\Users\Naught\AppData\Local\Programs\Python\Python310\lib\site-packages\torch\autograd\grad_mode.py", line 27, in decorate_context
return func(*args, **kwargs)
TypeError: LatentDiffusion.on_train_batch_start() missing 1 required positional argument: 'dataloader_idx'
Epoch 0: 0%| | 0/2020 [00:52<?, ?it/s]

---

Not solved yet.Googleing...................

[andreae293]

i think you are starting from a messy environment
probably the correct packages are not installed

This is what i did just a moment ago on a fresh new windows installation with 32GB RAM:
-install anaconda3
-install git on windows from there
-add git executable path to environment variables (on my pc located on C:\Program Files\Git\bin )

-download this repo somewhere on your pc
-place model.ckpt in the right folder
-create regularization and training folder and put images there

-open anaconda prompt and navigate to this repo

cd Dreambooth-Stable-Diffusion-cpu-main
conda env create -f environment.yaml
conda activate ldm
python main.py --base configs/stable-diffusion/v1-finetune_unfrozen.yaml -t --actual_resume ./models/ldm/stable-diffusion-v1/model.ckpt -n dog_testing  --data_root ./training/ --reg_data_root ./regularization --class_word dog --logdir ./logs/

and its working

be advsided that 32GB are not enough and you will get memory error
reserve something like 10GB or more for virtual ram

edit: also check that
conda env create -f environment.yaml
does not return an error

[78Alpha]

We were both trying to run from a newer environment. However, it involves manually editing a lot of things to get it "working".

Even altering the site packages themselves.

[andreae293]

We were both trying to run from a newer environment. However, it involves manually editing a lot of things to get it "working".

Even altering the site packages themselves.

@78Alpha
unfortunately i too ran in this kind of problems all the time when using any ML-related project
i am under the impression that those environments breaks really easily: installing a different version of a package, even an updated one can break the environment.
all i can suggest you is to clean the old anaconda environment, even manually deleting the "ldm" environment folder and recreate it with the steps i wrote

@cyruschan360
im glad that its working for you
did you have to reserve any memory to virtual RAM?

[78Alpha]

@NaughtDZ

TypeError: LatentDiffusion.on_train_batch_start() missing 1 required positional argument: 'dataloader_idx'

main.py, ~456, on_train_batch_end
def on_train_batch_end(self, trainer, pl_module, outputs, batch, batch_idx, dataloader_idx=0):
and ddpm.py ~509 on_train_batch_start
def on_train_batch_start(self, batch, batch_idx, dataloader_idx=0):

They are set to 0 because it doesn't even use those to begin with as far as I can tell, but I keep them in just in case.

[78Alpha]

@andreae293

Seems to be a linux thing, as @NaughtDZ got the Automatic1111 extension working on CPU by the looks
AUTOMATIC1111/stable-diffusion-webui#4386

Also referenced AUTOMATIC1111/stable-diffusion-webui#2002 (comment)
Seems to always kill after 1 step no matter what in WSL2 at least

But native windows worked fine
AUTOMATIC1111/stable-diffusion-webui#2002 (comment)

[andreae293]

that pull request looks like it has an even faster cpu training than this repo (33s/it vs 50s/it), i'm going to give it a try

[78Alpha]

The speed varies, I am running at fastest 47s/it on a Ryzen 7 3700x, but it does work, no doubt. Model saving still has some issues but that can be done after.