Instructions to use Lonuhbow/makeout2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use Lonuhbow/makeout2 with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("krea/Krea-2-Turbo", dtype=torch.bfloat16, device_map="cuda") pipe.load_lora_weights("Lonuhbow/makeout2") prompt = "Makeout2" image = pipe(prompt).images[0] - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- Draw Things
- DiffusionBee
Upload folder using huggingface_hub
Browse files
.gitattributes
CHANGED
|
@@ -43,3 +43,4 @@ custom_watch/step-1400-comparisons.jpg filter=lfs diff=lfs merge=lfs -text
|
|
| 43 |
custom_watch/step-1600-comparisons.jpg filter=lfs diff=lfs merge=lfs -text
|
| 44 |
custom_watch/step-1800-comparisons.jpg filter=lfs diff=lfs merge=lfs -text
|
| 45 |
custom_watch/step-2000-comparisons.jpg filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
| 43 |
custom_watch/step-1600-comparisons.jpg filter=lfs diff=lfs merge=lfs -text
|
| 44 |
custom_watch/step-1800-comparisons.jpg filter=lfs diff=lfs merge=lfs -text
|
| 45 |
custom_watch/step-2000-comparisons.jpg filter=lfs diff=lfs merge=lfs -text
|
| 46 |
+
training_metrics/progress.png filter=lfs diff=lfs merge=lfs -text
|
training_metrics/checkpoint_summary.md
ADDED
|
@@ -0,0 +1,20 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Training diagnostics by checkpoint
|
| 2 |
+
|
| 3 |
+
The training metrics help flag instability. Training loss alone does not measure identity or prompt flexibility.
|
| 4 |
+
|
| 5 |
+
| Step | Loss (50-step mean) | Loss variability | Gradient norm | Clipped steps (last 50) | LR | Seconds/step (last 50) | Identity score |
|
| 6 |
+
| ---: | ---: | ---: | ---: | ---: | ---: | ---: | ---: |
|
| 7 |
+
| 200 | 0.1735 | 0.1879 | 0.04248 | 1/50 | 0.00025 | 1.46 | — |
|
| 8 |
+
| 400 | 0.1607 | 0.1969 | 0.01868 | 0/50 | 0.00025 | 1.48 | — |
|
| 9 |
+
| 600 | 0.1811 | 0.201 | 0.0108 | 0/50 | 0.00025 | 1.48 | — |
|
| 10 |
+
| 800 | 0.1838 | 0.2403 | 0.03931 | 0/50 | 0.00025 | 1.56 | — |
|
| 11 |
+
| 1000 | 0.1189 | 0.1193 | 0.02673 | 0/50 | 0.00025 | 1.53 | — |
|
| 12 |
+
| 1200 | 0.135 | 0.1671 | 0.02515 | 0/50 | 0.00025 | 1.54 | — |
|
| 13 |
+
| 1400 | 0.1378 | 0.1488 | 0.0188 | 0/50 | 0.00025 | 1.54 | — |
|
| 14 |
+
| 1600 | 0.1643 | 0.2016 | 0.06348 | 0/50 | 0.00025 | 1.43 | — |
|
| 15 |
+
| 1800 | 0.1242 | 0.1471 | 0.01855 | 0/50 | 0.00025 | 1.52 | — |
|
| 16 |
+
| 2000 | 0.1982 | 0.2427 | 0.1553 | 0/50 | 0.00025 | 1.47 | — |
|
| 17 |
+
|
| 18 |
+
Identity Watch results were unavailable.
|
| 19 |
+
|
| 20 |
+
GPU peak is recorded per step in training_metrics.jsonl. The seconds/step window includes checkpoint saves when present.
|
training_metrics/progress.png
ADDED
|
Git LFS Details
|
training_metrics/training_metrics.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|