no option to include negative prompts in python CLI help?

[sarahspak]

Ran python -m python_coreml_stable_diffusion.pipeline --help but am not seeing an option for negative_prompts, though based on this issue, it seems it should be available unless I'm misunderstanding something?

usage: pipeline.py [-h] --prompt PROMPT -i I -o O [--seed SEED] [--model-version MODEL_VERSION] [--compute-unit {ALL,CPU_AND_GPU,CPU_ONLY,CPU_AND_NE}]
                   [--scheduler {DDIM,DPMSolverMultistep,EulerAncestralDiscrete,EulerDiscrete,LMSDiscrete,PNDM}] [--num-inference-steps NUM_INFERENCE_STEPS] [--guidance-scale GUIDANCE_SCALE]
                   [--controlnet [CONTROLNET ...]] [--controlnet-inputs [CONTROLNET_INPUTS ...]]

options:
  -h, --help            show this help message and exit
  --prompt PROMPT       The text prompt to be used for text-to-image generation.
  -i I                  Path to input directory with the .mlpackage files generated by python_coreml_stable_diffusion.torch2coreml
  -o O
  --seed SEED, -s SEED  Random seed to be able to reproduce results
  --model-version MODEL_VERSION
                        The pre-trained model checkpoint and configuration to restore. For available versions: https://huggingface.co/models?search=stable-diffusion
  --compute-unit {ALL,CPU_AND_GPU,CPU_ONLY,CPU_AND_NE}
                        The compute units to be used when executing Core ML models. Options: ('ALL', 'CPU_AND_GPU', 'CPU_ONLY', 'CPU_AND_NE')
  --scheduler {DDIM,DPMSolverMultistep,EulerAncestralDiscrete,EulerDiscrete,LMSDiscrete,PNDM}
                        The scheduler to use for running the reverse diffusion process. If not specified, the default scheduler from the diffusers pipeline is utilized
  --num-inference-steps NUM_INFERENCE_STEPS
                        The number of iterations the unet model will be executed throughout the reverse diffusion process
  --guidance-scale GUIDANCE_SCALE
                        Controls the influence of the text prompt on sampling process (0=random images)
  --controlnet [CONTROLNET ...]
                        Enables ControlNet and use control-unet instead of unet for additional inputs. For Multi-Controlnet, provide the model names separated by spaces.
  --controlnet-inputs [CONTROLNET_INPUTS ...]
                        Image paths for ControlNet inputs. Please enter images corresponding to each controlnet provided at --controlnet option in same order.

Version: pip freeze | grep "diffusion"
Returns

python-coreml-stable-diffusion @ git+https://github.com/apple/ml-stable-diffusion@5d2744e38297b01662b8bdfb41e899ac98036d8b

[atiorh]

Hello @sarahspak, the issue you referenced points to a PR that added negative prompts to the Swift CLI only. If you would like to add negative prompts to the Python CLI as well, feel free to put up a PR for that. Thanks for bringing this up!

[sarahspak]

got it - happy to do so!
I did notice an argument for [negative_prompt](https://github.com/apple/ml-stable-diffusion/blob/ce0be82717016f9a0bbdc6a09202860808203e7e/python_coreml_stable_diffusion/pipeline.py#L112) in the pipeline.py file. Was this feature not fully implemented?

[atiorh]

Just pushed the change to unblock you, please open another issue if you hit any problems 🙏

[sarahspak]

thank you @atiorh!