Trouble running on M3 Pro

[deKeijzer]

Hi there,

Just letting you know that it does not work out of the box on a M3 pro. The error I get is not of much use.

It might be helpful if --num_class_images could become a variable in the code, just for testing purposes. Now it takes quite a while to even get to the point of the error.

Bests

---------------------------------------------------------------------------
CalledProcessError                        Traceback (most recent call last)
File <timed exec>:1

File /opt/homebrew/lib/python3.9/site-packages/py_dreambooth/trainer.py:124, in LocalTrainer.fit(self, model, dataset)
    118 command = model.make_command(self.config_path)
    120 log_or_print(
    121     f"The model training has begun.\n'max_train_steps' is set to {max_train_steps}.",
    122     self.logger,
    123 )
--> 124 _ = subprocess.run(shlex.split(command), check=True)
    125 log_or_print("The model training has ended.", self.logger)
    127 predictor = LocalPredictor(model, self.output_dir, self.logger)

File /opt/homebrew/Cellar/[email protected]/3.9.18_2/Frameworks/Python.framework/Versions/3.9/lib/python3.9/subprocess.py:528, in run(input, capture_output, timeout, check, *popenargs, **kwargs)
    526     retcode = process.poll()
    527     if check and retcode:
--> 528         raise CalledProcessError(retcode, process.args,
    529                                  output=stdout, stderr=stderr)
    530 return CompletedProcess(process.args, retcode, stdout, stderr)

CalledProcessError: Command '['accelerate', 'launch', '/opt/homebrew/lib/python3.9/site-packages/py_dreambooth/scripts/train/train_sd_dreambooth.py', '--pretrained_model_name_or_path', 'stabilityai/stable-diffusion-2-1-base', '--instance_data_dir', 'data_preproc', '--instance_prompt', 'a photo of brian person', '--num_class_images', '200', '--output_dir', 'models', '--resolution', '768', '--train_batch_size', '1', '--learning_rate', '2e-06', '--lr_scheduler', 'constant', '--lr_warmup_steps', '0', '--validation_prompt', 'a photo of brian person with Eiffel Tower in the background', '--compress_output', 'False', '--with_prior_preservation', 'True', '--prior_loss_weight', '1.0', '--class_prompt', 'a photo of person', '--train_text_encoder', 'True', '--max_train_steps', '800', '--gradient_accumulation_steps', '1', '--gradient_checkpointing', 'True', '--use_8bit_adam', 'True', '--enable_xformers_memory_efficient_attention', 'True', '--mixed_precision', 'fp16', '--set_grads_to_none', 'True', '--report_to', 'tensorboard']' returned non-zero exit status 1.