Stable Diffusion Example within a Jupyter Notebook using Mac A1 Chip
Review the Setup post here
In Jupyter, navigate to the stable-diffusion folder and create a new notebook. Let’s check our Python Version.
import sys
sys.path.append(".")
sys.versionThe output from this cell should be something like this.
3.10.6 (main, Oct 24 2022, 11:04:07) [Clang 12.0.0 ]Now let’s check our OS.
import platform
platform.platform()
The output from this cell should be something like this
macOS-13.0-arm64-arm-64bitNow let’s just call the txt2img.py file with a prompt as a test.
%run scripts/txt2img.py --prompt "a red juicy apple floating in outer space, like a planet" --n_samples 1 --n_iter 1 --plms
Here is the output. See bottom of the page for troubleshooting.
NOTE: Redirects are currently not supported in Windows or MacOs.
Global seed set to 42
Loading model from models/ldm/stable-diffusion-v1/model.ckpt
Global Step: 470000
LatentDiffusion: Running in eps-prediction mode
DiffusionWrapper has 859.52 M params.
making attention of type 'vanilla' with 512 in_channels
Working with z of shape (1, 4, 32, 32) = 4096 dimensions.
making attention of type 'vanilla' with 512 in_channels
Some weights of the model checkpoint at openai/clip-vit-large-patch14 were not used when initializing CLIPTextModel: ['vision_model.encoder.layers.8.mlp.fc1.weight', 'vision_model.encoder.layers.23.mlp.fc1.weight', 'vision_model.encoder.layers.5.self_attn.q_proj.weight',
---snipped----
'vision_model.encoder.layers.23.self_attn.out_proj.bias', 'vision_model.encoder.layers.14.layer_norm1.bias', 'vision_model.encoder.layers.15.mlp.fc2.weight', 'vision_model.encoder.layers.3.layer_norm2.weight', 'vision_model.encoder.layers.21.layer_norm2.bias']
- This IS expected if you are initializing CLIPTextModel from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).
- This IS NOT expected if you are initializing CLIPTextModel from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).
Sampling: 0%| | 0/1 [00:00<?, ?it/s]
data: 0%| | 0/1 [00:00<?, ?it/s][A
Data shape for PLMS sampling is (1, 4, 64, 64)
Running PLMS Sampling with 50 timesteps
PLMS Sampler: 0%| | 0/50 [00:00<?, ?it/s][A[A
PLMS Sampler: 2%|▏ | 1/50 [00:08<06:36, 8.10s/it][A[A
PLMS Sampler: 4%|▍ | 2/50 [00:08<03:03, 3.83s/it][A[A
PLMS Sampler: 6%|▌ | 3/50 [00:09<01:53, 2.41s/it][A[A
PLMS Sampler: 8%|▊ | 4/50 [00:10<01:20, 1.75s/it][A[A
PLMS Sampler: 10%|█ | 5/50 [00:11<01:02, 1.39s/it][A[A
PLMS Sampler: 12%|█▏ | 6/50 [00:11<00:51, 1.16s/it][A[A
PLMS Sampler: 14%|█▍ | 7/50 [00:12<00:44, 1.03s/it][A[A
PLMS Sampler: 16%|█▌ | 8/50 [00:13<00:39, 1.06it/s][A[A
PLMS Sampler: 18%|█▊ | 9/50 [00:14<00:36, 1.13it/s][A[A
PLMS Sampler: 20%|██ | 10/50 [00:14<00:33, 1.20it/s][A[A
PLMS Sampler: 22%|██▏ | 11/50 [00:15<00:31, 1.24it/s][A[A
PLMS Sampler: 24%|██▍ | 12/50 [00:16<00:30, 1.26it/s][A[A
PLMS Sampler: 26%|██▌ | 13/50 [00:17<00:28, 1.30it/s][A[A
PLMS Sampler: 28%|██▊ | 14/50 [00:17<00:27, 1.30it/s][A[A
PLMS Sampler: 30%|███ | 15/50 [00:18<00:26, 1.31it/s][A[A
PLMS Sampler: 32%|███▏ | 16/50 [00:19<00:26, 1.31it/s][A[A
PLMS Sampler: 34%|███▍ | 17/50 [00:20<00:25, 1.31it/s][A[A
PLMS Sampler: 36%|███▌ | 18/50 [00:20<00:24, 1.31it/s][A[A
PLMS Sampler: 38%|███▊ | 19/50 [00:21<00:23, 1.31it/s][A[A
PLMS Sampler: 40%|████ | 20/50 [00:22<00:22, 1.34it/s][A[A
PLMS Sampler: 42%|████▏ | 21/50 [00:23<00:21, 1.33it/s][A[A
PLMS Sampler: 44%|████▍ | 22/50 [00:23<00:21, 1.33it/s][A[A
PLMS Sampler: 46%|████▌ | 23/50 [00:24<00:20, 1.33it/s][A[A
PLMS Sampler: 48%|████▊ | 24/50 [00:25<00:19, 1.35it/s][A[A
PLMS Sampler: 50%|█████ | 25/50 [00:26<00:18, 1.34it/s][A[A
PLMS Sampler: 52%|█████▏ | 26/50 [00:26<00:17, 1.34it/s][A[A
PLMS Sampler: 54%|█████▍ | 27/50 [00:27<00:17, 1.33it/s][A[A
PLMS Sampler: 56%|█████▌ | 28/50 [00:28<00:16, 1.35it/s][A[A
PLMS Sampler: 58%|█████▊ | 29/50 [00:29<00:15, 1.37it/s][A[A
PLMS Sampler: 60%|██████ | 30/50 [00:29<00:14, 1.35it/s][A[A
PLMS Sampler: 62%|██████▏ | 31/50 [00:30<00:14, 1.34it/s][A[A
PLMS Sampler: 64%|██████▍ | 32/50 [00:31<00:13, 1.34it/s][A[A
PLMS Sampler: 66%|██████▌ | 33/50 [00:32<00:12, 1.33it/s][A[A
PLMS Sampler: 68%|██████▊ | 34/50 [00:32<00:12, 1.33it/s][A[A
PLMS Sampler: 70%|███████ | 35/50 [00:33<00:11, 1.33it/s][A[A
PLMS Sampler: 72%|███████▏ | 36/50 [00:34<00:10, 1.34it/s][A[A
PLMS Sampler: 74%|███████▍ | 37/50 [00:35<00:09, 1.33it/s][A[A
PLMS Sampler: 76%|███████▌ | 38/50 [00:35<00:09, 1.32it/s][A[A
PLMS Sampler: 78%|███████▊ | 39/50 [00:36<00:08, 1.32it/s][A[A
PLMS Sampler: 80%|████████ | 40/50 [00:37<00:07, 1.32it/s][A[A
PLMS Sampler: 82%|████████▏ | 41/50 [00:38<00:06, 1.32it/s][A[A
PLMS Sampler: 84%|████████▍ | 42/50 [00:38<00:05, 1.34it/s][A[A
PLMS Sampler: 86%|████████▌ | 43/50 [00:39<00:05, 1.34it/s][A[A
PLMS Sampler: 88%|████████▊ | 44/50 [00:40<00:04, 1.33it/s][A[A
PLMS Sampler: 90%|█████████ | 45/50 [00:41<00:03, 1.36it/s][A[A
PLMS Sampler: 92%|█████████▏| 46/50 [00:41<00:02, 1.35it/s][A[A
PLMS Sampler: 94%|█████████▍| 47/50 [00:42<00:02, 1.36it/s][A[A
PLMS Sampler: 96%|█████████▌| 48/50 [00:43<00:01, 1.37it/s][A[A
PLMS Sampler: 98%|█████████▊| 49/50 [00:43<00:00, 1.36it/s][A[A
PLMS Sampler: 100%|██████████| 50/50 [00:44<00:00, 1.12it/s][A[A
data: 100%|██████████| 1/1 [00:49<00:00, 49.09s/it][A
Sampling: 100%|██████████| 1/1 [00:49<00:00, 49.09s/it]
Your samples are ready and waiting for you here:
outputs/txt2img-samples
Enjoy.
Navigate to the outputs folder above and there should be an image file starting grid-000x.png
Troubleshooting
- Read the errors provided and install/update missing librabries
- Google or the error
- Compare output from the first two cells above, how does your Python and OS compare? You may need to update your OS or Python?
Leave a comment